1. Introduction
The Tibetan Plateau is the highest plateau on Earth with a mean altitude in excess of 4000 m. Snow occurs frequently during winter and spring across this region according to satellite-retrieved, reanalyzed, and observed in situ snow depth and snow-cover data (Pu et al. 2007; Chu et al. 2014). The average number of annual snowfall days is 60, with 45% in spring, 28% in winter, 22% in autumn, and 5% in summer; a maximum of 154 annual snowfall days has been recorded on the central Tibetan Plateau (Chu et al. 2017). Snow has a considerable impact on the surface energy balance and water cycle via albedo and hydrological effects (Xu and Dirmeyer 2013), and plays a vital role in modulating regional atmospheric circulation patterns (W. Li et al. 2018).
Albedo is a key parameter in land surface models; it strongly depends on the microphysical properties of snow (i.e., particle size, shape, and liquid water content) (Aoki et al. 2003) and is highly sensitive to fresh snow, snow age (Abolafia-Rosenzweig et al. 2022), and snow depth (Wang et al. 2020). Albedo is also affected by the deposition of light-absorbing particles (He et al. 2018; X. Li et al. 2018; He 2022), which amplify snow albedo feedback loops (He et al. 2021) and severely modify the duration and evolution of snow cover. Albedo is a key source of uncertainty in the simulation of snow states in land surface models (Chen et al. 2014; Abolafia-Rosenzweig et al. 2021). Snow albedo feedback is important because the loss of snow cover reduces the amount of solar radiation reflected back into space, which in turn amplifies climate forcing and further accelerates snowmelt, enhancing future warming (Walton et al. 2017; Winter et al. 2017). However, it is difficult to accurately represent snow albedo feedback in current regional climate models (Minder et al. 2016); this is a result of challenges in the physically based consideration of all influencing factors in the parameterization of snow cover and albedo, accurate modeling of snow-cover dynamics, and capturing the complex interactions with other climate processes such as cloud formation, which yield large biases in snow-cover and albedo simulations.
Various snow albedo parameterization schemes have been developed, evaluated, and improved through the explicit involvement of additional influencing factors and optimization of complex snow processes (Wang et al. 2013; Oaida et al. 2015; Liu et al. 2021, 2022a,b; Abolafia-Rosenzweig et al. 2022; Ström et al. 2022; Hao et al. 2023). Considering the effect of snow states on albedo parameterization and the fact that wet snow has a larger snow albedo decay rate compared with dry snow, surface skin temperature has been selected as a proxy for snow state (wet or dry) (Wang et al. 2013).
In addition, optimizing the parameters in the snow albedo scheme of the multiphysics Noah land surface model has substantially improved the accuracy of snow albedo estimates, particularly during albedo decay periods, increasing the broadband snow albedo Nash–Sutcliffe Efficiency from −2.03 to 0.66 (Abolafia-Rosenzweig et al. 2022). A further snow albedo scheme has been derived that depends on the ambient temperature, snow depth, and concentration of light-absorbing particles (Ström et al. 2022). Albedo and snow-cover simulations have also been improved through the consideration of various types of nonspherical snow grain shapes and different mixing states of snow and light-absorbing particles, advancing the understanding of the roles of both snow grain shape and impurity mixing state in land surface processes (Hao et al. 2023).
The Noah land surface model involves one-dimensional water and heat exchange and only a single snowpack layer (Ek et al. 2003). Although the Noah model was developed several decades ago with simple and robust parameterizations, and it performs worse than current state-of-the-art Noah land surface model with multiphysics options (Noah-MP) and sophisticated Community Land Model (CLM) (Liu et al. 2019), its high computational efficiency means that it is widely used in numerical weather prediction and simulation over high mountainous regions such as the Tibetan Plateau (Meng et al. 2018; Zhang et al. 2019, 2021; Ji et al. 2022). However, owing to the overparameterization of snow albedo on the Tibetan Plateau in the Noah scheme, the state-of-the-art Weather Research and Forecasting (WRF) Model coupled with Noah shows an evident cold temperature bias (Meng et al. 2018; Liu et al. 2019), and snow-related properties, such as albedo and snow cover, are poorly characterized by the model (Malik et al. 2014; Minder et al. 2016; Meng et al. 2018). These previous results suggest that there is a necessity to improve the Noah snow albedo scheme using in situ observations over mountainous regions such as the Tibetan Plateau, where snowfall accumulation and melting differs from that in regions with lower topography. This difference is mainly determined by climate, topography, and altitude. In addition to the seasonal characteristics of snow cover over the Tibetan Plateau, differences lie in the rapid accumulation and slow melting of snow across high mountainous regions compared with the slow accumulation and rapid melting of shallow, patchy, and short-lived snow on the interior of the plateau. Furthermore, explicitly considering snow depth and optimizing parameters in a modified Noah albedo scheme using in situ observations and satellite-retrieved data can yield great improvements in the simulation of snow albedo effects and surface energy budgets during snowfall and snowmelt processes; this methodology has already been preliminarily validated against sparse in situ observations for eight snowfall events (Liu et al. 2021) and the modified Noah albedo scheme shows comparable performance to a sophisticated CLM for an additional snowfall event (Liu et al. 2022b).
The Tibetan Plateau has a heterogeneous underlying surface and a sparsity and uneven distribution of meteorological stations, with most in situ stations located in relatively flat valleys and rarely on steep slopes or ridges. Moreover, meteorological stations are rarely located in the uninhabited areas of the northwestern Tibetan Plateau that experience harsh climatic conditions. The albedo effect of snow varies with topography and land-cover type. Therefore, the lack of representativeness of in situ observations brings limitations in their use to verify the performance of our previously improved snow albedo scheme across the entire Tibetan Plateau region (Liu et al. 2021, 2022b). Accurate simulations of regional snow albedo effects require further evaluation and improvement in the treatment of snow cover and albedo in regional climate models. Here, satellite-retrieved products [i.e., albedo from the Moderate Resolution Imaging Spectroradiometer (MODIS) and snow cover from the Interactive Multisensor Snow and Ice Mapping System (IMS)] during eight snowfall events from winter 2018 to spring 2019 were selected in this study to verify the applicability of our previously improved snow albedo scheme for albedo and snow-cover simulation over the entire Tibetan Plateau. The eight snowfall events are the same as those used in our previous work (Liu et al. 2021). However, this study focused on the skill scores of the model validated by satellite-retrieved products of albedo and snow cover across the entire Tibetan Plateau; this is in contrast to our previous work where sparse ground observations of snow depth, snow water equivalent, air temperature, turbulent heat, and water vapor exchanges were used to evaluate the performance of the model.
2. Data and methods
a. Ground observations and satellite-retrieved products
The hourly air temperature and precipitation from 254 national stations of the China Meteorological Administration (CMA) were used to determine the daily snowfall accumulation over the Tibetan Plateau. The criterion for judging the occurrence of snowfall was precipitation when the air temperature was lower than freezing point. Then, ground observations of daily snowfall accumulation were used to select and characterize the intensity and spatial distribution of snowfall events from winter 2018 to spring 2019. The freezing point was selected as the threshold in the ground observations of snowfall accumulation because the freezing point was used for snow and rain recognition in the Noah land surface model; this unifies snow and rain recognition between observations and simulations.
To fill the gaps between ground observation stations in depopulated regions of the northwestern Tibetan Plateau, satellite-retrieved albedo and snow-cover products were selected. Snow-cover maps were produced using IMS observations, which are derived from products including in situ observations and satellite imagery. The daily GeoTIFF formatted snow cover from winter 2018 to spring 2019 was used with spatial resolutions of 1 and 4 km (U.S. National Ice Center 2008). The IMS snow-cover product is considered valid at 0000 coordinated universal time (UTC) and is available from https://nsidc.org/data/g02156/versions/1.
b. Overview of the eight snowfall events
Snowfall, snow cover, and snow albedo are interconnected components of Earth’s cryosphere. Snow cover represents the accumulation of snowfall over a specific region and period. It is determined by the intensity and frequency of snowfall events, and it also depends greatly on the drivers, timing, magnitude, and spatial variability of snowmelt. Snow albedo is highly related to snow cover and changes sharply during snowfall and snowmelt processes. High snow albedo would result in low surface net radiation and low surface temperature, which in turn influences the rate and timing of snowmelt and finally the snow cover. In this study, snowfall events occurring from winter 2018 to spring 2019 were selected to investigate the applicability of our new snow albedo scheme to the improvement of snow-cover simulation. According to the national standard grades of snowfall from the China Meteorological Standardization Network, daily snowfall of 2.5–5.0, 5.0–10.0, and 10.0–20.0 mm is classed as moderate, heavy, and snowstorm, respectively (http://www.cmastd.cn/standardView.jspx?id=479). In total, eight snowfall events ranging in intensity from moderate to snowstorm were chosen. The CMA ground observations of the maximum daily snowfall accumulation at each CMA station during each snowfall event over the Tibetan Plateau are shown in Fig. 1. The snowfall dates, maximum accumulated snow, and maximum daily snowfall accumulation are shown in Table 1. Heavy snowfall mainly occurred on the eastern and southern Tibetan Plateau. The maximum daily snowfall accumulation reached 73.7 mm in February 2019, occurring at Nyalam on the southern slope of the Himalayas.
China Meteorological Administration ground observations of maximum daily snowfall accumulation (unit: mm) during each snowfall event across the Tibetan Plateau.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0083.1
Overview of the eight snowfall events from China Meteorological Administration ground observations. The maximum snowfall accumulation is obtained from a single station.
From the IMS snow-cover maps shown in Fig. 2, it can be seen that snow covered approximately half of the plateau during the period spanned by the eight snowfall events. The spatial distribution of snow cover changed little over this time in the northwestern and Himalayan regions, but varied to a greater extent across the other regions. Evidently, snow-cover observations existed in the northern and northwestern regions where CMA ground snow observations did not.
Interactive Multisensor Snow and Ice Mapping System snow-cover data for the eight snowfall events across the Tibetan Plateau.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0083.1
c. WRF Model and configuration of experiments
Simulations of albedo and snow cover were produced using the Advanced Research WRF (WRF-ARW) Model (Skamarock et al. 2008). Figure 3 depicts the two-way interactively nested domains within the model. The outer domain (d01) was used to simulate synoptic-scale atmospheric conditions over the Tibetan Plateau and surrounding regions with a 5-km horizontal grid spacing, while the inner domain (d02) was configured with a 1-km horizontal grid spacing. There were 35 unevenly spaced vertical layers stretching up to 50 hPa for both domains. The Thompson microphysics scheme (Thompson et al. 2008), unified Noah land surface scheme (Tewari et al. 2004), revised MM5 surface layer scheme (Jiménez et al. 2012), Dudhia shortwave radiation scheme (Dudhia 1989), RRTM longwave radiation scheme (Mlawer et al. 1997), YSU boundary layer scheme (Hong et al. 2006), and Kain–Fritsch cumulus scheme (Kain 2004) were chosen for the experiments. The cumulus scheme was turned off in the inner domain.
WRF Model outer (d01) and inner (colored rectangles denote d02) domains with shaded topographical height. The black, red, green, white, and purple rectangles represent the d02 of Ev1, Ev2, Ev4, Ev5, and Ev6, respectively.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0083.1
The improved albedo scheme is based on the principle that snow albedo is equal to MODIS-retrieved albedo in the case of deep snow (snow depth > 0.2 m) and snow albedo is equal to fresh snow albedo on days with snowfall. Fresh snow and snow-free albedo values were derived from MODIS-retrieved albedo. More detailed information regarding the improvement of the snow albedo scheme and its evaluation against MODIS-retrieved albedo for one snowfall event, and against in situ observations for all eight snowfall events, can be found in our previous work (Liu et al. 2021, 2022b).
Two experiments were conducted for each snowfall event: one implemented the default Noah snow albedo scheme and the other implemented our improved snow albedo scheme. The fifth-generation European Centre for Medium-range Weather Forecasts (ECMWF) reanalysis dataset (ERA5) with 3-h temporal and 0.25° spatial resolution provided the initial and boundary conditions for the experiments. The ERA5 dataset is freely available online from the website https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era5. The experiments were run from one day before the onset of each snowfall event until one day after the snowfall ceased. The simulation results were output at 3-h intervals, and the first day was used for model spinup. The detailed experimental configuration of run time and grid number of domains for each snowfall event is shown in Table 2.
Experimental configuration of run time and grid number of domains.
d. Simulation verification and validation
Two-by-two contingency table of a binary snow-cover flag.
The EH describes the overall accuracy of the model including the occurrence and nonoccurrence of snow; PO indicates the miss rate whereby the model fails to simulate positive snow cover; NH indicates the false-alarm rate whereby the model incorrectly simulates positive snow cover. Both TS and ETS are statistical verification measures to assess the simulation accuracy for binary snow cover; TS takes into account three components: correctly simulated positive snow cover (a in Table 3), failure to simulate positive snow cover (c in Table 3), and incorrectly simulated positive snow cover (b in Table 3). It reflects the proportion of snow occurring in both the simulation and observation. As TS cannot distinguish the source of simulation error, it is often considered together with the miss and false-alarm rates. However, TS depends on the climate frequency of the snowfall events. For low probability snowfall events, the score value of TS is low. The ETS absorbs the advantages of TS and additionally considers correctly simulated negative snow cover (d in Table 3). It is an improvement on TS, reducing the impact of random hit probability on scoring, making the skill score of the model fairer. Higher values for TS, ETS, and EH accompanied by lower values for PO and NH indicate a more skillful simulation. The maximum score of 1 for TS, ETS, and EH indicates a perfect simulation without misses (PO = 0) and false alarms (NH = 0).
3. Results
a. Spatial pattern of albedo
Snow mainly appeared in the eastern, central, and southern regions, and stably dominated in the northwestern and Himalayan regions (Fig. 2). Albedo varied sharply during snowfall and subsequent snowmelt events. The mean spatial pattern of the observed and simulated albedo during each snowfall event over the Tibetan Plateau is shown in Fig. 4. For each of the eight snowfall events, albedo > 0.4 was retrieved from MODIS where IMS defined snow cover. The satellite-retrieved albedo was >0.6 in the eastern, central, and southern regions, and much larger (>0.8) in the northwestern and Himalayan regions. The WRF Model in which the default Noah albedo scheme was applied produced a wide extent of large albedo values with remarkable overestimates across the whole Tibetan Plateau, except for small areas of the northwestern and Himalayan regions, where the observed albedo exceeded 0.8. The WRF Model in which our improved Noah albedo scheme was applied significantly outperformed the WRF default scheme model in terms of albedo estimates; it greatly alleviated albedo overestimation, and accurately reproduced not only the spatial pattern over the Tibetan Plateau but also the large spatial difference of albedo between 0.2 and 0.8 in the northwestern and Himalayan regions on the edge of the Tibetan Plateau (Fig. 4).
Mean albedo for each snowfall event (Ev1–Ev8) over the Tibetan Plateau. The terms “default” and “new” refer to the WRF Model applying the default and improved Noah albedo scheme, respectively.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0083.1
b. Skill of the model in terms of albedo estimates
The spatial distribution of albedo bias and relative bias between satellite observations and model simulations over the whole Tibetan Plateau is shown in Fig. 5, and that of albedo RMSE is shown in Fig. 6. Owing to the large overestimation of snow-cover extent and snow-covered surface albedo, the WRF Model applying the default Noah albedo scheme gave a rather large RMSE and large positive bias and relative bias of albedo. The apparent alleviation of the albedo overestimation resulted in the overall reduction of the albedo bias, relative bias, as well as the RMSE estimated by the model using the improved albedo scheme across the Tibetan Plateau; there were apparent improvements relative to the RMSE from WRF using the default scheme (Figs. 5 and 6).
Spatial distribution of albedo bias and relative bias for each snowfall event over the Tibetan Plateau. For the definition of “default” and “new,” see Fig. 4.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0083.1
Spatial distribution of albedo RMSE and RMSE difference (new minus default) for each snowfall event over the Tibetan Plateau. For the definition of “default” and “new,” see Fig. 4.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0083.1
For regions with snow cover, the albedo RMSE throughout snowfall events depends disproportionately on not only the differences in parameters, such as the maximum snow albedo or fresh snow albedo, but also the snow albedo decay rate during the snowmelt processes. The maximum snow albedo in the default scheme is also constant depending only on the land surface vegetation type from the lookup table of the model; for example, the maximum snow albedo on the snow and ice surface is set to 0.82 from the vegetation lookup table, while the fresh snow albedo in the improved scheme is a constant 0.79 from the MODIS mean fresh snow albedo on the Tibetan Plateau. Besides, the default Noah albedo scheme gives a small and constant snow albedo decay rate, which contradicts the rapid melting of snow on the interior regions of the Tibetan Plateau. The improved albedo scheme characterized the rapid and variable melting of snow. It is a more physically based scheme and suitable for the Tibetan Plateau. These are potential reasons explaining the albedo RMSE decrease across the Tibetan Plateau using the improved albedo scheme when evaluating against equivalent MODIS product sources.
The temporal evolution of regional mean albedo, and the bias, relative bias, RMSE, and correlation coefficient between the satellite-retrieved and WRF-estimated albedo over the Tibetan Plateau is shown in Fig. 7. Albedo increases to its maximum value when fresh snow falls and then decreases to a low value when snow melts. The temporal evolution of albedo is therefore often associated with the snow state. During the eight snowfall events, the observed regional mean albedo ranged from 0.32 to 0.53 and exhibited an inverted U-shaped temporal evolution from a low albedo of 0.32, to a high snow albedo of 0.53 during snowfall, followed by albedo decay during snowmelt. This inverted U-shaped temporal pattern of albedo was driven by the snowfall and subsequent snowmelt processes; this is because albedo increases with snow depth and snow-covered area during fresh snowfall, and decreases with snow age or liquid water content in snow during snowmelt.
Daily regional mean albedo, and albedo bias, relative bias, RMSE, and correlation coefficient (corr) between MODIS retrievals (MOD) and model estimates with 5-km grid spacing over the whole Tibetan Plateau. The x axis shows the first date of Ev1–Ev8. For the definition of “default” and “new,” see Fig. 4.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0083.1
In this study, the mean albedo across the eight snowfall events was 0.44. Compared with the MODIS albedo product, the WRF Model in which the default Noah albedo scheme was applied gave a consistent overestimation, with an albedo range of 0.52–0.68 (mean: 0.62). There was no clear daily variation pattern of albedo estimated using the default scheme with failure in reproducing the inverted U-shaped daily evolution of albedo retrieved from MODIS. The WRF Model in which our improved Noah albedo scheme was applied reduced the overestimated albedo by 0.13–0.27 (mean: 0.18), resulting in similar albedo values to the observations. In addition, an inverted U-shaped temporal pattern was reproduced by the model applying our improved scheme, especially for Ev1, Ev2, Ev4, and Ev5, with the explicit consideration of snow depth and optimization of snow-age-related parameters serving as potential reasons. Although capturing the inverted U-shaped daily variation is not a novel albedo scheme improvement, as the Biosphere–Atmosphere Transfer Scheme (BATS) and Canadian Land Surface Scheme (CLASS) applied in the widely used Noah-MP already capture this variation (Minder et al. 2016; Abolafia-Rosenzweig et al. 2022; Li et al. 2022), the improvement for the Noah albedo scheme indeed represents significant progress. These data confirmed the success of the improved scheme in capturing albedo evolution from large snow albedo to albedo decay over the whole Tibetan Plateau; this pattern is not captured by the default Noah scheme.
From the albedo bias and relative bias statistics, the default albedo scheme gave maximum and mean overestimations of 0.28 and 0.18, respectively, and the relative overestimation was 58%–190% (mean: 104%). Our improved albedo scheme gave a bias range of ±0.08 and mean bias close to zero, and the relative bias decreased to −2%–82% (mean: 34%). On average, the relative bias using our improved scheme decreased by 70% when compared with that using the default scheme. From the albedo RMSE and correlation coefficient analysis, the default scheme gave a large RMSE of 0.26–0.35 (mean: 0.29) and small correlation coefficient of 0.15–0.52 (mean: 0.36). Using our improved scheme, the albedo RMSE was reduced by 0.01–0.16 (mean: 0.07), and the albedo correlation coefficient was markedly increased by 0.03–0.39 (mean: 0.13), meeting the 99% significance level for a two-tailed t test (Fig. 7).
The improved albedo scheme only modified albedo over the snow-covered land surface. In addition to the overall evaluation of albedo schemes over the whole Tibetan Plateau, a comparison of albedo over the snow-covered land surface of the Tibetan Plateau was also conducted to further assess the performance of our improved snow albedo scheme. Here, the snow-covered region was defined as both model snow-cover estimates and satellite snow products pointing to snow presence. The temporal evolution of regional mean albedo, and the related bias analysis between the satellite-retrieved and WRF-estimated albedo over the snow-covered Tibetan Plateau region is shown in Fig. 8. It is evident that the inverted U-shaped daily variation of albedo generated from snow processes is successfully reproduced by the model applying our improved albedo scheme. The default albedo scheme gave a larger snow-covered surface albedo (mean: 0.7) when compared with the MODIS albedo (mean: 0.58), leading to a mean positive albedo bias of 0.12 and relative bias of 31%. In contrast, the improved scheme yielded a slightly smaller mean surface albedo (0.53) when compared with the MODIS albedo, resulting in a bias close to zero, with a mean negative albedo bias of –0.05 and relative bias of –4%. Moreover, the mean snow-covered surface albedo correlation coefficient was markedly increased from the default scheme (0.19) to the improved scheme (0.34), meeting the 99% significance level for a two-tailed t test. However, both schemes gave the same mean RMSE of 0.21 implying that both positive and negative deviations existed in daily grided albedo statistics over the snow-covered region. Furthermore, the apparent improvement of snow albedo estimates using the new albedo scheme appeared in the deep snow-covered regions (snow depth > 0.2 m), where the mean albedo of 0.64 from MODIS and the new scheme and 0.75 from the default scheme (not shown). For the shallow snow-covered regions (snow depth < 0.2 m), however, neither of the both albedo schemes performed well with the mean overestimates of 0.12 for the default scheme and the mean underestimates of –0.11 for the new scheme (not shown).
As in Fig. 7, but for the snow-covered land region over the Tibetan Plateau.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0083.1
Albedo on snow-free land surfaces is determined by surface vegetation type and vegetation coverage in WRF Model. Vegetation information is important in snow-free albedo estimation on the Tibetan Plateau because of the largely heterogeneous distribution of vegetation type across this region. Furthermore, we compared the satellite-retrieved and WRF-estimated albedo for snow-free land surfaces. Compared with the mean MODIS albedo of 0.16, the model gave a large mean albedo of 0.26 with a mean bias of 0.1, relative bias of 92%, and RMSE of 0.13. Such the large bias of albedo on snow-free land surfaces is the main source of albedo simulation deviation for the improved scheme during snowfall events across the entire Tibetan Plateau, although there is relatively low albedo bias on the snow-covered areas. However, the significant source of albedo simulation deviation for the default scheme includes the large albedo bias not only on the snow-free land surfaces but also on the snow-covered regions, which could be concluded from the large mean bias and relative bias of albedo on both snow-covered and snow-free areas. Potential reasons for the significant bias between the model-estimated and satellite-retrieved albedo for snow-free land surfaces include a mismatch in the static land surface type in WRF compared with that in MODIS in 2018/19, or inaccuracies in the snow-free albedo calculated from the lookup table of the model and vegetation shadow fraction. For example, the snow-free albedo was 0.38 for barren and sparsely vegetated surfaces in the WRF Model, while the satellite-retrieved albedo was <0.3 for the same grided region.
The temporal evolution of regional mean albedo and the skill scores between the satellite-retrieved and WRF-estimated albedo at 1-km grid spacing over d02 are displayed in Fig. 9. The model applying our improved scheme outperformed that applying the default scheme over the eastern Tibetan Plateau in characterizing the daily variation in albedo from large snow albedo to albedo decay during snowfall and snowmelt processes, substantially alleviating albedo overestimation, decreasing the simulated bias and RMSE, increasing the correlation coefficient, and providing a mean simulated albedo value more similar to observations. On average, the model applying our improved scheme decreased the relative bias of albedo by 14%–128% (mean improvement: 56%), decreased the albedo RMSE by 0.06, and increased the correlation coefficient by 0.14, with a maximum improvement of 0.39 (Fig. 9). Over the eastern Tibetan Plateau, the improved scheme performed better than the default scheme during relatively weak snowfall events (i.e., Ev2 and Ev4–Ev6); however, it performed worse during heavy to blizzard events (i.e., Ev1), although the improved scheme reproduced the rapid increase in albedo during snowfall and rapid albedo decay during snowmelt of Ev1 (Fig. 9). Furthermore, our improved WRF simulation at 1-km grid spacing showed an equivalent improvement to that of albedo estimates at 5-km grid spacing over the eastern Tibetan Plateau.
As in Fig. 7, but for model estimates with 1-km grid spacing over the nested domain. The x axis shows the first date of Ev1, Ev2, and Ev4–Ev6.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0083.1
c. Skill of model in terms of snow-cover estimates
Over snow-covered regions, the surface albedo is determined by snow-related variables, such as snow cover, snow depth, and snow age, and in turn has a slight impact on snow cover through adjusting the surface energy budget. Comparing the observations of snow depth and daily snowfall accumulation on 254 CMA ground stations from Liu et al. (2021) and WRF simulations at 1- and 5-km grid spacing over the Tibetan Plateau, it showed that WRF significantly overestimated both snow variables, with the improved albedo scheme alleviating the snow depth overestimation (not shown). For the evaluation of snow-cover estimates, the regional mean snow cover of the IMS retrievals and model estimates is displayed in Fig. 10. Compared with the IMS snow-cover product, both the default and improved models significantly overestimated mean snow cover. This was a result of the overestimation of solid precipitation in the WRF Model. For the default scheme, the large overestimation of snow cover would lead to highly biased albedo over the snow-covered region because of the explicit consideration of snow cover in this scheme. However, an overestimation of snow cover of equal magnitude was caused by the solid precipitation overestimation in the improved model. For thick snow regions (snow depth > 0.2 m), the surface albedo was similar to the snow albedo, with a large snow albedo decay rate from the improved scheme. For shallow snow regions (snow depth < 0.2 m), in addition to the large snow albedo decay rate, the surface albedo was highly dependent on the low underlying surface albedo of 0.19 from the improved scheme (Liu et al. 2022b). This contributes to the substantial decrease in regional mean albedo overestimation (Figs. 7–10).
Daily regional mean snow cover of IMS retrievals and model estimates at (a) 1-km grid spacing over the eastern Tibetan Plateau, with the x axis denoting the first date of Ev1, Ev2, and Ev4–Ev6, and (b) 5-km grid spacing over the whole Tibetan Plateau, with the x axis denoting the first date of Ev1–Ev8. For the definition of “default” and “new,” see Fig. 4.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0083.1
Compared with the IMS snow-cover product, the WRF Model applying our improved scheme at 1-km grid-spacing configuration reduced the overestimation of snow cover over the eastern Tibetan Plateau, except for Ev1 (Fig. 10a), and consistently reduced the overestimation of snow cover for all snowfall events over the whole Tibetan Plateau when the model was configured with 5-km grid spacing (Fig. 10b). The snow-cover overestimation at different grid spacings was alleviated to a limited extent, with a regional mean reduction of 0.05.
To comprehensively evaluate the effect of our improved albedo scheme on snow-cover estimates across the entire Tibetan Plateau, and evaluate the snow-cover simulation skill of the model, the common skill scores (i.e., TS, ETS, PO, NH, and EH) and correlation coefficient were calculated (see Fig. 11). Evidently, our improved scheme consistently outperformed the default scheme in terms of snow-cover simulations, with 0.03-higher mean TS and ETS skill scores, indicating a higher simulation accuracy for the binary snow cover. The default scheme gave a notably larger extent of snow cover compared with IMS retrievals. In terms of statistical verifications of the miss rate (PO) and false-alarm rate (NH), the default scheme had a very small PO of 0.01–0.09 (mean: 0.02) and rather large NH (mean: 0.45). Owing to the reduction in snow-cover overestimation, our improved scheme consistently decreased the NH by 0.03 and increased the overall accuracy by 0.04 (EH), although there was also an evident increase in the PO. In addition, compared with the default scheme simulation, the spatial correlation coefficient of snow cover between the improved scheme simulation and IMS product across the entire Tibetan Plateau increased by 0.05, meeting the 99% significance level for a two-tailed t test (Fig. 11). These statistical verifications highlight the positive effect of the improved snow albedo scheme on snow-cover estimates over the Tibetan Plateau resulting in relatively high accuracy and a low false-alarm rate.
Daily skill scores for the model in terms of threat score (TS), equitable threat score (ETS), point over, not hit, and efficient hit for snow-cover estimates, and the correlation coefficient (corr) between IMS-retrieved snow cover and model estimates with 5-km grid spacing over the whole Tibetan Plateau. The x axis shows the first date of Ev1–Ev8. For the definition of “default” and “new,” see Fig. 4.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0083.1
The above skill scores were also used to evaluate the accuracy of snow-cover simulation at 1-km grid spacing during the five snowfall events that occurred over the eastern Tibetan Plateau (see Fig. 12). Compared with the improvement in the WRF simulation at 5-km grid spacing over the whole Tibetan Plateau, an equivalent improvement was observed for these skill scores of snow-cover estimates using the model applying our improved scheme in 1-km grid-spacing simulations over the eastern Tibetan Plateau, although a large miss rate (PO) was calculated (Figs. 11 and 12). In addition, equivalent skill scores from snow-cover simulations at 1- and 5-km grid spacing over the eastern Tibetan Plateau were obtained.
As in Fig. 11, but for model estimates with 1-km grid spacing over the nested domain. The x axis shows the first date of Ev1, Ev2, and Ev4–Ev6.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0083.1
d. The linkage between high-biased snow cover and low-biased albedo estimates
WRF using either the default or the improved albedo scheme generated the similar spatial distribution of snow cover, depth. and the past 24-h variation across the Tibetan Plateau. WRF using the default albedo scheme apparently overestimated snow cover and as a result overestimated albedo at both 1- and 5-km grid spacing. However, WRF using the improved albedo scheme overestimated snow cover with albedo underestimation to a certain extent at both 1- and 5-km grid spacing.
To explore the potential explanations of high-biased snow cover and low-biased albedo, the spatial distribution of IMS snow cover and WRF using the improved albedo scheme estimates of snow cover, depth, and its variations over the past 24 h across the Tibetan Plateau is displayed in Fig. 13. For each snowfall events, a large proportion of snow cover presented to be shallow (snow depth < 0.2 m) in a melting state (negative snow depth variation) in regions where only WRF classified snow. The rapid snow albedo decay related with snow age of the improved scheme resulted in rapid reduction of snow albedo. The surface albedo further decreased significantly due to the overparameterization of the impact of the shallow snow-covered underlying surface on albedo. These were the potential reasons of apparent albedo underestimation for the improved albedo scheme across the whole Tibetan Plateau under the high-biased snow cover.
Interactive Multisensor Snow and Ice Mapping System (IMS) snow cover and WRF estimates across the Tibetan Plateau, and WRF-estimated snow depth and its variations over the past 24 h in regions where only WRF classifies snow for the eight snowfall events at 0000 UTC on the ends of the simulations using the new albedo scheme. For snow-cover flags, the gray (1) denotes that neither IMS nor WRF classify snow, the yellow (2) denotes that both IMS and WRF classify snow, the orange (3) denotes that only IMS classifies snow, and the blue (4) denotes that only WRF classifies snow.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0083.1
Furthermore, the spatial distribution of IMS snow cover, MODIS-retrieved albedo, and WRF using the improved albedo scheme estimates of albedo, snow cover and depth and its variations over the past 24 h over the nested domains is displayed in Fig. 14. Clearly, areas of high MODIS albedo were not completely consistent with areas where IMS classified snow. For example, there are areas of high MODIS albedo (>0.7) corresponding to no snow-covered areas classified by IMS (snow cover < 40%), especially on the eastern edge of the Tibetan Plateau. This was perhaps due to the cloud pollution that misled MODIS in discrimination of snow. This would cause MODIS-retrieved surface albedo to be unrealistically high, that is, MODIS excessively retrieved surface albedo. Additionally, WRF-estimated albedo using the improved scheme < 0.2 appeared in some areas where WRF estimated full snow cover that was both shallow and actively melting. Such the low WRF-estimated albedo was lower than the WRF background albedo from the model’s lookup table. This may result in low-biased regional mean albedo simulated by WRF applying the improved scheme. These above were the potential explanations that albedo was underestimated by WRF using the improved scheme across the eastern Tibetan Plateau where the high-biased snow cover located.
As in Fig. 13, but for model estimates with 1-km grid spacing on different dates over the nested domain with the distribution of MODIS-retrieved and WRF-estimated albedo using the improved scheme.
Citation: Monthly Weather Review 152, 3; 10.1175/MWR-D-23-0083.1
4. Summary and discussion
Using MODIS albedo products and assuming WRF simulated snow depth in previous control experiment as the “true value,” we obtained the improved albedo scheme that did not consider snow cover (Liu et al. 2022b). In this study, the performance of the WRF Model applying the improved albedo scheme in relation to albedo and snow-cover estimates was assessed against the MODIS albedo and IMS snow-cover products during eight snowfall events from winter 2018 to spring 2019 across the entire Tibetan Plateau. We showed that the WRF Model applying our improved albedo scheme significantly outperformed that applying the default Noah albedo scheme in terms of estimating albedo, reproducing large spatial differences in albedo (0.2–0.8) in the northwestern and Himalayan regions on the edge of the Tibetan Plateau, and successfully characterizing the spatial distribution of albedo across the entire plateau, with a mean improvement in the spatial correlation coefficient of 0.13. In addition, the improved scheme successfully characterized the inverted U-shaped temporal pattern of albedo during snowfall events. This was attributed to the explicit consideration of snow depth and native optimization of snow-age parameters in this scheme. Over the entire Tibetan Plateau, our improved scheme greatly alleviated the overestimation of albedo (overestimation reduction of 0.13–0.27; mean = 0.18), decreased the albedo bias to a range of ±0.08 with a relative bias improvement of 70%, and provided albedo values similar to true satellite-retrieved values.
The snowfall simulation is determined by the microphysics and cumulus schemes of the model, and is slightly impacted by the land surface energy balance. The reduction of albedo overestimation contributes to the alleviation of snow-cover overestimation across the Tibetan Plateau through the alteration of the surface energy balance. The default Noah albedo scheme results in a considerably larger extent of snow cover over the whole Tibetan Plateau when compared with IMS retrievals, which results in a very low miss rate (mean PO: 0.02) and rather large false-alarm rate (mean NH: 0.45). Owing to the alleviation of snow-cover overestimation, our improved scheme consistently decreases the false-alarm rate by 0.03, and increases the overall accuracy and spatial correlation coefficient by 0.04 and 0.05, respectively. Overall, these improvements compensate for the negative impact of the evident increase in miss rate, ultimately leading to increases of 0.03 in the comprehensive skill scores of TS and ETS, and improving the accuracy of snow-cover estimates over the Tibetan Plateau. Furthermore, the limited improvement of snow-cover estimates using the improved albedo scheme is possibly caused by the new treatment of snow albedo, including the explicit consideration of snow depth and large albedo decay rates over the Tibetan Plateau. This new treatment of snow albedo would affect the snow states and eventually the snow-cover extent through changing the land surface energy balance.
Qin et al. (2006) indicated that the snow is shallow, patchy, and short-lived over some of the Tibetan Plateau, with thick snow (tens of centimeters to several meters) occurring in mountainous regions and shallow snow (only a few centimeters) in the interior. Previous literature indicates that snow depth only plays a significant role in impacting snow albedo for shallow snow-covered regions (Wiscombe and Warren 1980). The importance of considering the impact of snow depth on snow albedo has been confirmed by Wang et al. (2020) in areas of shallow snow cover on the Tibetan Plateau, and ignorance of the dependence of snow albedo on snow depth in current snow schemes could potentially cause large model bias across the Tibetan Plateau. In this study, our previously improved snow albedo scheme performs better than the default Noah scheme in terms of snow-cover and albedo estimates. The improved albedo scheme showed significant improvements over the deep snow-covered regions, with less improvements over the shallow snow-covered regions; this can largely be attributed to parameter optimization related to snow albedo decay in the improved scheme. Explicitly considering the impact of snow depth on snow albedo in the improved albedo scheme under a single snowpack layer in Noah performs terribly over the shallow snow-covered regions. This can be attributed to the overparameterization of the impact of snow depth on albedo, which should be modified further.
The new albedo scheme is sensitive to snow depth in shallow snow regions not to snow layering schemes, which will lead to similar estimates of surface albedo when the new scheme is applied to different models with different snow layering schemes. However, Cristea et al. (2022) and many other studies (Dutra et al. 2012; Augas et al. 2020) have shown that snow surface temperature evolution influenced heavily by albedo and the resulting snow evolution can vary between models that represent snow layers differently, with single-layer models often having the large amount of thermal inertia. Therefore, it is necessary to further consider the snow layering configurations especially the top-layer thickness in the new albedo scheme in order to function well in models with different snow layering schemes.
In addition, it has been shown that large positive biases in albedo obtained from the Noah scheme result from substantial overpredictions of subpixel snow cover over snowy pixels, with a maximum overestimation of 0.4 occurring over midlatitude regions with complex and mountainous terrain (Minder et al. 2016). Modifications to areal snow depletion thresholds in Noah produce a more realistic snow albedo, and the reasonable treatment of snow albedo decay leads to substantial enhancements in the timing and magnitude of peak snowfall accumulation and snow-cover extent (Livneh et al. 2010). This is because the default Noah albedo scheme has a high dependence on snow cover. However, our results illustrate that positive biases in snow cover are inconsistent with positive albedo biases, a conclusion drawn from some periods in our improved albedo scheme exhibiting snow-cover overestimation accompanied by a negative albedo bias. The improved scheme optimizes the parameters related with the snow albedo decay, and overly parameterizes the influence of snow depth on albedo. The combination of MODIS excessive retrieval of surface albedo due to cloud pollution in snow-free areas and the underestimation of albedo under the shallow snow in a melting state are the potential reasons of a negative albedo bias for the improved scheme. In addition, the improved scheme shows the limited improvement on snow-cover overestimation through the surface energy balance. This cannot fundamentally reduce the high-biased snow-cover simulation to which the microphysical processes provide the main source of simulation errors.
Due to the additional influence of underlying land surface, the albedo decay of shallow snow is more complex than deep snow. Accurately parameterizing albedo in shallow snow areas presents a challenge. In this study, the improved albedo scheme does not distinguish between deep and shallow snow and overparameterizes the effect of snow depth on albedo with albedo underestimation over the shallow and actively melting snow. This would have significant impacts on the local rates of snowmelt and associated hydrology, in addition to the surface energy and mass fluxes. For example, albedo reduction accelerates when solar irradiance penetrates through the shallow snowpacks and the prescribed low albedo of the underlying surface increases absorption of solar irradiance. Snow melts rapidly and a large amount of snowmeltwater would infiltrate into the soil, increasing the modeled soil moisture and affecting the soil moisture flux and the terrestrial water storage variation. Besides, rapid snowmelt contributes significantly to the surface and the underground runoff, which would introduce large uncertainty in simulating streamflow and hydrologic extremes (Sun et al. 2019). To avoid albedo overparameterization and improve the underestimated albedo simulated by WRF using the improved scheme in shallow snow areas, on the one hand, we may find breakthroughs in optimizing the parameters related to snow albedo decay considering the different snow albedo decay curves of deep and shallow snow. On the other hand, we should reconsider the impact of snow depth on surface albedo to avoid overparameterization using the true snow-free background albedo and additionally considering more influencing factors of albedo. It has been confirmed to provide reasonable estimates of shallow snow-covered surface albedo during the melting of snow for models with more influencing factors such as near-surface air temperature, snowpack age, snow depth and density (Amaral et al. 2017).
For different grid spacing configurations (1 and 5 km), our improved scheme yields equivalent improvements, whether for albedo or snow-cover estimates, over the eastern Tibetan Plateau. However, a partial error remains after the improvement of the snow albedo scheme. The primary error source is related to the possibility that the observed fresh snow albedo is highly variable, whereas it is treated as a constant (Abolafia-Rosenzweig et al. 2022). During and after periods of snowfall, refinements in the snow scheme and more accurate specification of temperature-dependent fresh snow albedo can improve the performance of the model in terms of air temperature simulation (Smirnova et al. 2016). Therefore, building on our previously improved albedo scheme, further improvement could be achieved from the aspect of temperature-dependent fresh snow albedo using in situ observations.
Considering snow–aerosol interactions in land surface models has been shown to be successful in realistically simulating snow albedo and energy balances. Realistic aerosol deposition in snow results in a 6% albedo reduction, producing radiative forcing of 16 W m−2; this causes a 0.84°C surface warming and snowpack reduction of 11 mm (Oaida et al. 2015). The magnitudes of snow albedo reduction and surface radiation forcing are mainly related to the impurity content; for example, radiative effects induced by black carbon range from 0.7 to 58.4 W m−2 over the Tibetan Plateau (He et al. 2018). These impurity effects on snow albedo are ignored in the commonly used Noah land surface model. Therefore, in future studies, it will be necessary to consider the influence of impurity deposition on snow albedo in numerical weather prediction models using the Noah scheme to provide more accurate land surface process information over the Tibetan Plateau.
This study confirms the success of our previously improved snow albedo scheme and shows that it has great applicability potential in the simulation of snow albedo effects over the Tibetan Plateau, greatly alleviating previous cold air biases (Liu et al. 2021, 2022b). However, the improvement in snow-cover estimates obtained from the WRF Model applying our improved scheme is notably smaller than the improvement in albedo estimates; the snow cover is still largely overestimated by the WRF Model applying the improved albedo scheme. This is possibly related to a combination of the continuous overestimation of snowfall, systematic model wet bias, and defective snow-dependent parameterization of snow cover in the model. Furthermore, the importance of topographic effects in parameterizing snow cover and improving albedo simulation has been highlighted over the Tibetan Plateau (Miao et al. 2022). Therefore, for model users and developers, continued efforts should be made on the one hand to improve snowfall simulations and on the other hand to further optimize albedo and snow-cover schemes through considering topographic effects and additional snow-related variables in land surface models in the future.
Acknowledgments.
This research was supported by the National Natural Science Foundation of China (Grants 42230610 and 42205077), the Second Tibetan Plateau Scientific Expedition and Research program (STEP) (2019QZKK0103), and the National Key Scientific and Technological Infrastructure project “Earth System Numerical Simulation Facility” (EarthLab). We acknowledge the data providers of MODIS spectral reflectance, IMS snow-cover products, ERA5 data, and CMA snow observation data. The authors are very grateful to the reviewers for their professional reviews and for offering abundant and constructive comments.
Data availability statement.
The MODIS spectral reflectance product used in this study is openly available online (https://lpdaac.usgs.gov/products/mod09gav006/). The IMS snow-cover flag is openly available online (https://nsidc.org/data/g02156/versions/1). The ERA5 is freely available from https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era5. The ground snow observation data are from the staff of the China Meteorological Administration.
REFERENCES
Abolafia-Rosenzweig, R., C. He, S. P. Burns, and F. Chen, 2021: Implementation and evaluation of a unified turbulence parameterization throughout the canopy and roughness sublayer in Noah-MP snow simulations. J. Adv. Model. Earth Syst., 13, e2021MS002665, https://doi.org/10.1029/2021MS002665.
Abolafia-Rosenzweig, R., C. He, S. McKenzie Skiles, F. Chen, and D. Gochis, 2022: Evaluation and optimization of snow albedo scheme in Noah‐MP land surface model using in situ spectral observations in the Colorado Rockies. J. Adv. Model. Earth Syst., 14, e2022MS003141, https://doi.org/10.1029/2022MS003141.
Amaral, T., C. P. Wake, J. E. Dibb, E. A. Burakowski, and M. Stampone, 2017: A simple model of snow albedo decay using observations from the Community Collaborative Rain, Hail, and Snow-Albedo (CoCoRaHS-Albedo) network. J. Glaciol., 63, 877–887, https://doi.org/10.1017/jog.2017.54.
Aoki, T., A. Hachikubo, and M. Hori, 2003: Effects of snow physical parameters on shortwave broadband albedos. J. Geophys. Res., 108, 4616, https://doi.org/10.1029/2003JD003506.
Augas, J., K. Abbasnezhadi, A. N. Rousseau, and M. Baraer, 2020: What is the trade-off between snowpack stratification and simulated snow water equivalent in a physically-based snow model? Water, 12, 3449, https://doi.org/10.3390/w12123449.
Chen, F., and Coauthors, 2014: Modeling seasonal snowpack evolution in the complex terrain and forested Colorado Headwaters region: A model intercomparison study. J. Geophys. Res. Atmos., 119, 13 795–13 819, https://doi.org/10.1002/2014JD022167.
Chu, D., H. Xie, P. Wang, J. Guo, L. Jia, Y. Qiu, and Z. Zheng, 2014: Snow cover variation over the Tibetan Plateau from MODIS and comparison with ground observations. J. Appl. Remote Sens., 8, 084690, https://doi.org/10.1117/1.JRS.8.084690.
Chu, D., Q. Luosang, Z. Yang, and Y. Yang, 2017: Spatiotemporal variations of snowfall days over the Tibetan Plateau from 1981 to 2010 (in Chinese). J. Appl. Meteor. Sci., 28, 292–305, https://doi.org/10.11898/1001-7313.20170304.
Cristea, N. C., A. Bennett, B. Nijssen, and J. D. Lundquist, 2022: When and where are multiple snow layers important for simulations of snow accumulation and melt? Water Resour. Res., 58, e2020WR028993, https://doi.org/10.1029/2020WR028993.
Dudhia, J., 1989: Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model. J. Atmos. Sci., 46, 3077–3107, https://doi.org/10.1175/1520-0469(1989)046<3077:NSOCOD>2.0.CO;2.
Dutra, E., P. Viterbo, P. M. A. Miranda, and G. Balsamo, 2012: Complexity of snow schemes in a climate model and its impact on surface energy and hydrology. J. Hydrometeor., 13, 521–538, https://doi.org/10.1175/JHM-D-11-072.1.
Ek, M. B., K. E. Mitchell, Y. Lin, E. Rogers, P. Grunmann, V. Koren, G. Gayno, and J. D. Tarpley, 2003: Implementation of Noah land surface model advances in the National Centers for Environmental Prediction operational mesoscale Eta model. J. Geophys. Res., 108, 8851, https://doi.org/10.1029/2002JD003296.
Hao, D., and Coauthors, 2023: Improving snow albedo modeling in the E3SM land model (version 2.0) and assessing its impacts on snow and surface fluxes over the Tibetan Plateau. Geosci. Model Dev., 16, 75–94, https://doi.org/10.5194/gmd-16-75-2023.
He, C., 2022: Modelling light-absorbing particle–snow–radiation interactions and impacts on snow albedo: Fundamentals, recent advances and future directions. Environ. Chem., 19, 296–311, https://doi.org/10.1071/EN22013.
He, C., M. G. Flanner, F. Chen, M. Barlage, K.-N. Liou, S. Kang, J. Ming, and Y. Qian, 2018: Black carbon-induced snow albedo reduction over the Tibetan Plateau: Uncertainties from snow grain shape and aerosol–snow mixing state based on an updated SNICAR model. Atmos. Chem. Phys., 18, 11 507–11 527, https://doi.org/10.5194/acp-18-11507-2018.
He, C., F. Chen, R. Abolafia‐Rosenzweig, K. Ikeda, C. Liu, and R. Rasmussen, 2021: What causes the unobserved early‐spring snowpack ablation in convection‐permitting WRF modeling over Utah Mountains? J. Geophys. Res. Atmos., 126, e2021JD035284, https://doi.org/10.1029/2021JD035284.
Helfrich, S., M. Li, C. Kongoli, L. Nagdimunov, and E. Rodriguez, 2018: Interactive Multisensor Snow and Ice Mapping System Version 3 (IMS V3). NOAA NESDIS Algorithm Theoretical Basis Doc., version 2.5, 74 pp., https://nsidc.org/sites/default/files/ims_v3_atbd_v25.pdf.
Hong, S.-Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318–2341, https://doi.org/10.1175/MWR3199.1.
Ji, H., Z. Nan, J. Hu, Y. Zhao, and Y. Zhang, 2022: On the spin-up strategy for spatial modeling of permafrost dynamics: A case study on the Qinghai-Tibet Plateau. J. Adv. Model. Earth Syst., 14, e2021MS002750, https://doi.org/10.1029/2021MS002750.
Jiménez, P. A., J. Dudhia, J. F. Gonzalez-Rouco, J. Navarro, J. P. Montávez, and E. García-Bustamante, 2012: A revised scheme for the WRF surface layer formulation. Mon. Wea. Rev., 140, 898–918, https://doi.org/10.1175/MWR-D-11-00056.1.
Kain, J. S., 2004: The Kain–Fritsch convective parameterization: An update. J. Appl. Meteor., 43, 170–181, https://doi.org/10.1175/1520-0450(2004)043<0170:TKCPAU>2.0.CO;2.
Li, Q., T. Yang, and L. Li, 2022: Quantitative assessment of the parameterization sensitivity of the WRF/Noah-MP model of snow dynamics in the Tianshan Mountains, Central Asia. Atmos. Res., 277, 106310, https://doi.org/10.1016/j.atmosres.2022.106310.
Li, W., W. Guo, B. Qiu, Y. Xue, P.-C. Hsu, and J. Wei, 2018: Influence of Tibetan Plateau snow cover on East Asian atmospheric circulation at medium-range time scales. Nat. Commun., 9, 4243, https://doi.org/10.1038/s41467-018-06762-5.
Li, X., and Coauthors, 2018: Light-absorbing impurities in a Southern Tibetan Plateau glacier: Variations and potential impact on snow albedo and radiative forcing. Atmos. Res., 200, 77–87, https://doi.org/10.1016/j.atmosres.2017.10.002.
Liang, S., 2001: Narrowband to broadband conversions of land surface albedo I: Algorithms. Remote Sens. Environ., 76, 213–238, https://doi.org/10.1016/S0034-4257(00)00205-4.
Liu, L., Y. Ma, M. Menenti, X. Zhang, and W. Ma, 2019: Evaluation of WRF modeling in relation to different land surface schemes and initial and boundary conditions: A snow event simulation over the Tibetan Plateau. J. Geophys. Res. Atmos., 124, 209–226, https://doi.org/10.1029/2018JD029208.
Liu, L., Y. Ma, M. Menenti, R. Su, N. Yao, and W. Ma, 2021: Improved parameterization of snow albedo in Noah coupled with weather research and forecasting: Applicability to snow estimates for the Tibetan Plateau. Hydrol. Earth Syst. Sci., 25, 4967–4981, https://doi.org/10.5194/hess-25-4967-2021.
Liu, L., M. Menenti, and Y. Ma, 2022a: Evaluation of albedo schemes in WRF coupled with Noah-MP on the Parlung No. 4 Glacier. Remote Sens., 14, 3934, https://doi.org/10.3390/rs14163934.
Liu, L., M. Menenti, Y. Ma, and W. Ma, 2022b: Improved parameterization of snow albedo in WRF + Noah: Methodology based on a severe snow event on the Tibetan Plateau. Adv. Atmos. Sci., 39, 1079–1102, https://doi.org/10.1007/s00376-022-1232-1.
Livneh, B., Y. Xia, K. E. Mitchell, M. B. Ek, and D. P. Lettenmaier, 2010: Noah LSM snow model diagnostics and enhancements. J. Hydrometeor., 11, 721–738, https://doi.org/10.1175/2009JHM1174.1.
Malik, M. J., R. van der Velde, Z. Vekerdy, and Z. Su, 2014: Improving modeled snow albedo estimates during the spring melt season. J. Geophys. Res. Atmos., 119, 7311–7331, https://doi.org/10.1002/2013JD021344.
Meng, X., and Coauthors, 2018: Simulated cold bias being improved by using MODIS time-varying albedo in the Tibetan Plateau in WRF model. Environ. Res. Lett., 13, 044028, https://doi.org/10.1088/1748-9326/aab44a.
Miao, X., W. Guo, B. Qiu, S. Lu, Y. Zhang, Y. Xue, and S. Sun, 2022: Accounting for topographic effects on snow cover fraction and surface albedo simulations over the Tibetan Plateau in winter. J. Adv. Model. Earth Syst., 14, e2022MS003035, https://doi.org/10.1029/2022MS003035.
Minder, J. R., T. W. Letcher, and S. M. Skiles, 2016: An evaluation of high‐resolution regional climate model simulations of snow cover and albedo over the Rocky Mountains, with implications for the simulated snow‐albedo feedback. J. Geophys. Res. Atmos., 121, 9069–9088, https://doi.org/10.1002/2016JD024995.
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102, 16 663–16 682, https://doi.org/10.1029/97JD00237.
Oaida, C. M., Y. Xue, M. G. Flanner, S. M. Skiles, F. De Sales, and T. H. Painter, 2015: Improving snow albedo processes in WRF/SSiB regional climate model to assess impact of dust and black carbon in snow on surface energy balance and hydrology over western U.S. J. Geophys. Res. Atmos., 120, 3228–3248, https://doi.org/10.1002/2014JD022444.
Pu, Z., L. Xu, and V. V. Salomonson, 2007: MODIS/Terra observed seasonal variations of snow cover over the Tibetan Plateau. Geophys. Res. Lett., 34, L06706, https://doi.org/10.1029/2007GL029262.
Qin, D., S. Liu, and P. Li, 2006: Snow cover distribution, variability, and response to climate change in western China. J. Climate, 19, 1820–1833, https://doi.org/10.1175/JCLI3694.1.
Skamarock, W. C., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN‐475+STR, 113 pp., https://doi.org/10.5065/D68S4MVH.
Smirnova, T. G., J. M. Brown, S. G. Benjamin, and J. S. Kenyon, 2016: Modifications to the Rapid Update Cycle Land Surface Model (RUC LSM) available in the Weather Research and Forecasting (WRF) Model. Mon. Wea. Rev., 144, 1851–1865, https://doi.org/10.1175/MWR-D-15-0198.1.
Ström, J., and Coauthors, 2022: Snow albedo and its sensitivity to changes in deposited light-absorbing particles estimated from ambient temperature and snow depth observations at a high-altitude site in the Himalaya. Elementa, 10, 00118, https://doi.org/10.1525/elementa.2021.00118.
Sun, N., H. Yan, M. S. Wigmosta, L. R. Leung, R. Skaggs, and Z. Hou, 2019: Regional snow parameters estimation for large‐domain hydrological applications in the western United States. J. Geophys. Res. Atmos., 124, 5296–5313, https://doi.org/10.1029/2018JD030140.
Tewari, M., and Coauthors, 2004: Implementation and verification of the unified Noah land surface model in the WRF model. 20th Conf. on Weather Analysis and Forecasting/16th Conf. on Numerical Weather Prediction, Seattle, WA, Amer. Meteor. Soc., 14.2a, https://ams.confex.com/ams/84Annual/techprogram/paper_69061.htm.
Thompson, G., P. R. Field, R. M. Rasmussen, and W. D. Hall, 2008: Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part II: Implementation of a new snow parameterization. Mon. Wea. Rev., 136, 5095–5115, https://doi.org/10.1175/2008MWR2387.1.
U.S. National Ice Center, 2008: IMS daily Northern Hemisphere snow and ice analysis at 1 km, 4 km, and 24 km resolutions, version 1. National Snow and Ice Data Center, accessed 10 February 2023, https://doi.org/10.7265/N52R3PMC.
Vermote, E., and R. Wolfe, 2015: MOD09GA MODIS/Terra surface reflectance daily L2G Global 1 km and 500 m SIN Grid V006. NASA EOSDIS Land Processes DAAC, accessed 3 March 2023, https://doi.org/10.5067/MODIS/MOD09GA.006.
Walton, D. B., A. Hall, N. Berg, M. Schwartz, and F. Sun, 2017: Incorporating snow albedo feedback into downscaled temperature and snow cover projections for California’s Sierra Nevada. J. Climate, 30, 1417–1438, https://doi.org/10.1175/JCLI-D-16-0168.1.
Wang, T., and Coauthors, 2013: Evaluation of an improved intermediate complexity snow scheme in the ORCHIDEE land surface model. J. Geophys. Res. Atmos., 118, 6064–6079, https://doi.org/10.1002/jgrd.50395.
Wang, W., and Coauthors, 2020: Characterizing surface albedo of shallow fresh snow and its importance for snow ablation on the interior of the Tibetan Plateau. J. Hydrometeor., 21, 815–827, https://doi.org/10.1175/JHM-D-19-0193.1.
Winter, K. J.-P. M., S. Kotlarski, S. C. Scherrer, and C. Schär, 2017: The Alpine snow-albedo feedback in regional climate models. Climate Dyn., 48, 1109–1124, https://doi.org/10.1007/s00382-016-3130-7.
Wiscombe, W. J., and S. G. Warren, 1980: A model for the spectral albedo of snow. I: Pure snow. J. Atmos. Sci., 37, 2712–2733, https://doi.org/10.1175/1520-0469(1980)037<2712:AMFTSA>2.0.CO;2.
Xu, L., and P. Dirmeyer, 2013: Snow–atmosphere coupling strength. Part II: Albedo effect versus hydrological effect. J. Hydrometeor., 14, 404–418, https://doi.org/10.1175/JHM-D-11-0103.1.
Zhang, G., Z. Nan, X. Wu, H. Ji, and S. Zhao, 2019: The role of winter warming in permafrost change over the Qinghai‐Tibet Plateau. Geophys. Res. Lett., 46, 11 261–11 269, https://doi.org/10.1029/2019GL084292.
Zhang, G., Z. Nan, L. Zhao, Y. Liang, and G. Cheng, 2021: Qinghai-Tibet Plateau wetting reduces permafrost thermal responses to climate warming. Earth Planet. Sci. Lett., 562, 116858, https://doi.org/10.1016/j.epsl.2021.116858.
