1. Introduction
Snow albedo—the ratio of the snow-reflected solar radiation to the total solar radiation incident on the snowpack—is generally higher than land cover such as soils and vegetation during most of snow accumulation seasons and thus snow albedo exerts important impacts on regional to global terrestrial water and energy budgets. Variations of snow albedo influence temperature changes, as well as water supply from the streamflow through snowpack evolution and melting (Barnett et al. 2005; Flanner et al. 2011; Qian et al. 2015; Skiles et al. 2018; Zhang et al. 2019). Warming temperature reduces the snow-cover area and therefore snow albedo, leading to increased surface radiative forcing and even higher land surface temperature, forming positive snow-albedo feedback in the climate system (Hall 2004; Thackeray and Fletcher 2016). The snow-albedo feedback influences surface hydrological processes such as soil moisture, snow-cover area, evapotranspiration, and runoff. Therefore, accurate representation and simulation of snow albedo processes in weather and climate models can reduce the uncertainty in hydrometeorological predictions, allowing stakeholders to make more informed decisions.
There are still limitations and substantial uncertainties in the representation of snow albedo processes within many land surface models (LSMs) coupled to regional and global weather and climate models, which consequently affect the estimation of land surface energy and water balances. For example, systematic snow albedo biases were found in one of the most widely used LSMs, the Noah LSM with multiparameterization options (Noah-MP) (Niu et al. 2011), which is a land component within the Weather Research and Forecasting (WRF) Model, the NOAA Unified Forecast System (UFS) model, and the National Water Model (NWM) among many others, as demonstrated by various studies (Abolafia-Rosenzweig et al. 2021, 2022a; Chen et al. 2014; He et al. 2019a, 2021; Liu et al. 2021, 2022; Wang et al. 2020; Xiao et al. 2021). The snow albedo in Noah-MP is represented using semiphysical or empirical functions (Verseghy 1991; Yang et al. 1997), which lack physical treatments of snow albedo response to the evolution of snow properties such as aging and metamorphism. This approach leads to an inconsistent treatment between snow albedo and other snowpack properties (He and Flanner 2020). To enhance the modeling of snow albedo evolution, it is necessary to have a comprehensive understanding of the underlying physical mechanisms that influence snow albedo. Subsequently, these processes need to be represented in a more physically realistic manner.
The albedo of snow is determined by a complex combination of multiple factors, such as snow depth, snowpack structures, the size and shape of snow grains, and the concentration of light-absorbing particles (LAPs) (Warren and Wiscombe 1980; Flanner et al. 2021). It has been known that snow albedo is affected by LAPs mainly in the visible band (Warren and Wiscombe 1980) and by grain size mainly in the near-infrared (NIR) band (Wiscombe and Warren 1980). Following the occurrence of snowfall, snow crystals experience fast alterations in their size and shape, exhibiting a tendency for snow grains to progressively increase in size over time (Colbeck 1982). The alteration in snow grain size influences the interaction between the snow surface and incoming solar radiation. Snow with larger grains has a lower albedo due to an increase in the pathlength traveled by photons (Warren 1982). The Snow, Ice, and Aerosol Radiative (SNICAR) model (Flanner et al. 2007, 2021) is one of the most widely used open-source snowpack radiative transfer models, which resolves the aforementioned physical processes and simulates snow albedo by considering snowpack properties such as grain size and shape, as well as environmental conditions including the presence of LAPs.
Simulating snow albedo using SNICAR has many advantages compared to the current semiphysical snow albedo schemes in Noah-MP, such as the Biosphere-Atmosphere Transfer Scheme (BATS) (Abolafia-Rosenzweig et al. 2022a; Yang et al. 1997). 1) The study conducted by Abolafia-Rosenzweig et al. (2022a) demonstrated the issue of using a constant parameter to represent the fresh snow albedo in NoahMP–BATS, which is typically adopted by empirical/semiphysical schemes, whereas the measured albedo of fresh snow exhibits significant variability, especially in the NIR band. In contrast, fresh snow albedo is influenced by several environmental conditions and physical processes in the SNICAR model, such as temperature, downward solar spectrum, snow grain size and shapes, LAPs within the snow, and the thickness and density of the snowpack. The SNICAR treatment is more physically realistic, as pointed out by Wang et al. (2020). 2) The inclusion of different snow grain shapes, such as spheres, spheroids, hexagonal plates, and Koch snowflakes, is necessary to depict the types of nonspherical snow grains that are commonly observed (Liou et al. 2014; He et al. 2018a, 2024; Robledano et al. 2023). This representation is currently absent in Noah-MP, while it has been included in SNICAR. 3) The representation of snow aging processes is empirical and incomplete in Noah-MP due to the absence of explicit inclusion of snow grain growth. Instead, the simulated snow grain size in SNICAR can be validated using in situ measurements and remote sensing products. This approach offers the advantage of requiring less arbitrary tuning of empirical snow aging parameters, which is needed by current Noah-MP snow albedo schemes. 4) SNICAR simulates the interaction between snow, aerosols, and radiation (Flanner et al. 2021; He et al. 2018a; Skiles and Painter 2019), encompassing three LAPs: black carbon (BC), organic carbon (OC), and dust. Additionally, the latest SNICAR coupled into the Community Land Model, version 5 (CLM5), and the DOE’s Energy Exascale Earth System Model (E3SM) Land Model (ELM) includes the internal mixing of BC and dust with snow grains (He et al. 2024; Hao et al. 2023). However, these treatments are either missing or not physically represented in Noah-MP. 5) SNICAR presents the effect of solar zenith angle on snow albedo (for direct radiation) physically, while Noah-MP parameterizes this effect semiempirically such as in BATS. 6) SNICAR computes vertical solar radiation absorption and heating for individual snow layers, which changes snow and soil temperature profiles but is missing in the current Noah-MP albedo schemes. 7) SNICAR has a hyperspectral calculation capability that is more accurate than narrowband calculations (Wang et al. 2022) and can be expanded to incorporate or compare with spectral radiation obtained by remote sensing, while current Noah-MP snow albedo schemes only use two (visible and NIR) bands.
While recent studies have included SNICAR into land surface models such as CLM and ELM in global climate models (Hao et al. 2023; He et al. 2024), there are several unique implications and potential applications of coupling SNICAR with Noah-MP LSM. 1) Noah-MP has a strong emphasis on surface hydrological processes, such as representing land surface processes in the WRF–Hydro/National Water Model for operational hydrological forecasting (Cosgrove et al. 2024). The accuracy of snow albedo processes on snowmelt magnitude and timing influences basin streamflow estimates, making it especially important for hydrological studies and water resource management (Sthapit et al. 2022; von Kaenel and Margulis 2024; Xiao et al. 2021). 2) Noah-MP is widely used in operational weather prediction models such as the NOAA Unified Forecast System (UFS) and regional climate models such as WRF Model applications across various regions and spatial scales (Powers et al. 2017; Rasmussen et al. 2023). The integration of SNICAR can be particularly beneficial for regions with significant snow cover or where snow processes have a large impact on local and regional climate and hydrology. 3) Noah-MP has a proven track record in land data assimilation (e.g., Kumar et al. 2008; Peters-Lidard et al. 2007), which improves its performance in operational settings and enables continuous model improvements based on the observational data. The coupled NoahMP–SNICAR model’s hyperspectral resolution could help to effectively combine snow surface reflectance information from remote sensing observations with land surface models, improving albedo prediction, land–atmosphere interactions, and compensating for shortcomings during periods when remote sensing observations are missing (Shao et al. 2020). 4) We also refactored the SNICAR codes based on the approach introduced with the recent release of refactored Noah-MP v5.0 (He et al. 2023a). The coupled NoahMP–SNICAR improves model code standards and data structures, increasing model modularity, interoperability, and application. The refactored version of SNICAR allows for easy coupling to different climate/weather/hydrology models, tailoring the model to specific research needs or operational requirements.
The overall objective of this study is to investigate whether a sophisticated physically based snow albedo scheme SNICAR coupled with the Noah-MP LSM can improve simulated snow albedo accuracy. The most recent SNICAR version includes various new features and enhancements, which account for snow grain shape and size, snow–aerosol–radiation interaction, and snow aging processes (section 2). We evaluate NaohMP–SNICAR modeled snow albedo using in situ spectral observations at three Rocky Mountain field stations used for Noah-MP snow albedo evaluations by Abolafia-Rosenzweig et al. (2022a) (section 3 and Table S1 in the online supplemental material). We first determine whether the coupled NoahMP–SNICAR model can accurately reproduce the observed mean and variability of snow albedo (section 4a). Then, we compare the snow albedo between the new NoahMP–SNICAR simulation and the default semiphysical NoahMP–BATS simulation. Finally, to quantify effects of the novel processes in the NoahMP–SNICAR model, we perform the process-level model sensitivity experiments to assess the effects of snow grain sizes, snow grain shapes, and LAPs on snow albedo and surface radiative forcing (section 4b). Section 5 discusses the potential uncertainties and future directions, and section 6 concludes the study.
2. Noah-MP LSM and its coupling with SNICAR
a. Noah-MP model description
Noah-MP (Niu et al. 2011; Yang et al. 2011) is one of the most widely used open-source community LSMs worldwide, which has been used in various research and operational models pertaining to weather, climate, and hydrology. The newest version of Noah-MP (version 5.0) (He et al. 2023b) has undergone recent code refactoring/modernization and incorporated contemporary FORTRAN code styles, data structures, and standards. This refactoring has significantly improved the model’s modularity, interoperability, and applicability (He et al. 2023a).
Noah-MP is featured as a multiparameterization LSM that enables users to combine different physics schemes for modeling land surface processes (Niu et al. 2011). The Noah-MP snow module has the capability to simulate a maximum of three snow layers, with the number of layers being dependent on snow depth. Noah-MP treats explicit snow layers when snow depth is larger than 0.025 m and implicitly represents a very shallow (<0.025 m) snow layer by combining it with the top soil layer in energy and water balance calculations. Snow layer temperature, snow depth, and snow water and ice contents are calculated based on snowpack water and energy balances. The model considers many key snow processes such as snow layer division and combination, liquid water holding within the snowpack, snow compaction, snowmelting and refreezing, frost and sublimation, and the interception of snow by vegetation canopy. The minimum thicknesses for each snow layer from top to bottom layers are 0.025, 0.025, and 0.1 m, and the maximum thicknesses for each snow layer are 0.05 m, 0.2 m, and no limit (He et al. 2023b). During each model time step, snow layers may change with snow depth due to new snowfall, compaction, and ablation. Therefore, for a multilayer snowpack, snow layers are adjusted by combining them based on the minimum layer thickness; layers are then subdivided if they exceed the maximum layer thickness threshold. Snow layer thickness combination and division processes are also used to update snow ice, snow liquid water, and snow temperature. The technical report by He et al. (2023b) provides a comprehensive description of the various aspects related to snowpack mass and energy processes.
Within Noah-MP, there exist two semiphysical snow albedo schemes, namely, the Canadian Land Surface Scheme (CLASS; Verseghy 1991) and BATS (Yang et al. 1997). The mathematical equations of the two snow albedo schemes are described in He et al. (2023b). Both schemes simulate snow albedo in the visible (300–700 nm) and NIR (700–5000 nm) bands under direct and diffuse radiation, but CLASS assumes the same snow albedo for direct and diffuse radiation as well as visible and NIR bands, which is physically unrealistic. Furthermore, both schemes do not explicitly simulate the evolution of snow properties (e.g., snow aging/metamorphism, grain size, and shape). A recent study (Abolafia-Rosenzweig et al. 2022a) has tried to optimize the BATS albedo parameters using in situ snow albedo measurements, which, however, still showed important remaining biases particularly for fresh snow albedo due to a lack of physical representation of relevant albedo processes.
b. NoahMP–SNICAR model coupling
In this study, we couple the refactored Noah-MP, version 5, with the latest version of SNICAR (CTSM Development Team 2022) that has recently been implemented into CLM5 (He et al. 2024).
1) Multiple physics options for SNICAR albedo calculations
The SNICAR scheme we implement into Noah-MP incorporates several key physical processes and updates following He et al. (2024): 1) two options for radiative transfer solvers, with one for a traditional tridiagonal matrix two-stream solver (Toon et al. 1989) and one for a new adding-doubling solver (Dang et al. 2019); 2) three options for ice optical properties (Flanner et al. 2021) using different ice refractive indices from Warren (1984), Warren and Brandt (2008), and Picard et al. (2016); 3) updated aerosol optical properties of BC and OC from Flanner et al. (2021); 4) six options of representative downward solar spectra for multiple atmospheric conditions (Flanner et al. 2021), including midlatitude winter, midlatitude summer, subarctic winter, subarctic summer, Summit Greenland, and high mountain; 5) four types of snow grain shapes including sphere, spheroid, hexagonal, and snowflake (He et al. 2017); 6) three dust types including Saharan dust, Colorado dust, and Greenland dust (Flanner et al. 2021); 7) two options for either internal or external mixing of dust (He et al. 2019b) or BC (He et al. 2017) with snow grains; and 8) two options for wavelength band setup, including 5-band and hyperspectral (480 band with 10-nm spectral resolution) capabilities. Both 5-band and 480-band spectral albedo and absorbed solar radiation flux in each snow layer and surface soil layer are averaged, weighted by the downward solar spectrum (from a lookup table), to two (visible and NIR) bands to be used in the Noah-MP surface energy flux calculations.
The required SNICAR input variables for snow albedo (radiative transfer) calculations include direct/diffuse radiation, surface downward solar spectrum, solar zenith angle (only for direct radiation), albedo of the surface underlying snowpack, vertical profiles of snow grain size, snow layer thickness, snow density, and mass concentrations of LAPs (BC, mineral dust, and OC), snow grain shape, and optical properties of ice and LAPs. The optical properties of ice and LAPs for each snow layer and spectral bands, including single-scattering albedo, mass extinction cross section, and asymmetry parameter, are archived as lookup table datasets generated in Flanner et al. (2021) and He et al. (2024). The outputs from SNICAR that are passed to the Noah-MP model are snow albedo and fraction of absorbed solar radiation flux for each snow layer and the top soil layer (Fig. 1).
Schematic diagram for the coupled NoahMP–SNICAR land surface model. Green color represents variables passed to SNICAR radiative transfer scheme. Red color represents variables passed to Noah-MP from SNICAR radiative transfer scheme. New processes in the Noah-MP model include LAPs within snow and snow grain growth and aging.
Citation: Journal of Hydrometeorology 26, 2; 10.1175/JHM-D-24-0082.1
2) Snow grain growth and aging processes
The evolution of snow effective grain size is represented by wet and dry snow aging processes in NoahMP–SNICAR, including liquid-water-induced metamorphism, dry snow metamorphism, refreezing of liquid water, and the addition of freshly fallen snow (Flanner et al. 2007; Lawrence et al. 2019). The liquid-water-induced metamorphism is parameterized based on previously measured grain growth rates under different liquid water contents (Brun 1989). The dry snow metamorphism is determined by snow temperature, temperature gradient, density, and initial snow grain size distribution based on a microphysical particle model that simulates diffusive vapor flux among collections of ice crystals with various size and interparticle spacing (Flanner and Zender 2006). This process reproduces the typical observed rapid snow aging and increased snow grain size under the conditions of combined warm snow, large temperature gradient, and low density. The effective radius of refrozen liquid water is set to 1000 μm (Oleson et al. 2013). The surface air temperature is a key factor in determining the grain size of freshly fallen snow. At temperatures below −30°C, a minimum of 54.5 μm (radius) is imposed (Lawrence et al. 2019). A limit is imposed on the maximum of 204.5 μm (radius) when the temperature exceeds 0°C. A linear ramp is employed within the temperature range between −30° and 0°C (Lawrence et al. 2019). These maximum and minimum limits are calibratable parameters. In our investigation, we discover that the evolution of grain size and snow albedo is sensitive to the minimum and maximum values of freshly fallen snow grain size. In section 3c(3), we have optimized these two parameters to match with the snow grain size acquired from a satellite product (section 3b). In situations where there is a nonzero snow mass but an explicit snow layer has not yet been established (i.e., snow depth < 0.025 m), the effective snow grain size is assigned as the effective radius of freshly fallen snow (Lawrence et al. 2019). Snow grain size is adjusted proportionally to snow layer thickness changes as a result of layer combination and division. When two snow layers are combined, the effective snow grain size is calculated as a mass-weighted mean of those of the two layers. When a snow layer is divided into two, the effective snow grain sizes of the two layers are assumed to be the same as that before layer division. Last, the effective snow grain size is limited to a range of 30–1500 μm, as this range covers the majority of snow grain size in reality and corresponds to the defined optical properties that are archived in lookup tables (Flanner et al. 2021; He et al. 2024).
3) Snow–aerosol–radiation interactions
Additionally, we implement a mass-conserving approach to account for the presence of LAPs within the snowpack, encompassing the mechanisms of atmospheric aerosol deposition onto the uppermost snow layer, aerosol mass reduction via interlayer meltwater drainage, and aerosol mass changes due to snow layer combination and subdivision (Flanner et al. 2007; Lawrence et al. 2019). The mass of LAPs within snow is adjusted proportionally to snow layer thickness changes as a result of layer combination and division. NoahMP–SNICAR tracks the mass of nine aerosol particle species within each snow layer including hydrophilic BC, hydrophobic BC, hydrophilic OC, hydrophobic OC, and mineral dust with five particle size bins (μm in diameter; Table S2): 0.1–1.0, 1.0–2.5, 2.5–5.0, 5.0–10.0, and 10.0–100.0 (Flanner et al. 2021). Each species exhibits distinct optical characteristics (Flanner et al. 2021; He et al. 2024) and meltwater removal efficiencies (Lawrence et al. 2019). Aerosol is moved/removed from one layer to the layer underneath when meltwater drains from each layer during the melting process. Meltwater removal efficiencies determine the rate of aerosol mass movement and removal within each snow layer because of meltwater. These parameters are contingent upon species, ranging from 0.01 to 0.2, and are derived from Conway et al. (1996). Identical parameter values were used in other land surface models (e.g., Flanner et al. 2007; Lawrence et al. 2019; Oaida et al. 2015).
4) Albedo and snowpack heating
The results simulated from NoahMP–SNICAR include the spectral snow albedo and the fraction of solar flux that is absorbed by each snow layer and the top soil layer. The spectral snow albedo is partitioned into visible and NIR bands by computing the spectrally weighted mean based on the surface downward solar spectra (Flanner et al. 2007). The layerwise snowpack heating due to snow and LAPs absorption of solar radiation in SNICAR is coupled with Noah-MP snow and soil temperature computations to alter the temperature for each snow layer and the underlying top soil layer.
3. Data and model experiments
a. In situ observations
In situ observations of surface albedo and snow depth data are obtained from three high-elevation locations, East River, Irwin, and Senator Beck, within the southern Rocky Mountains in the state of Colorado, United States (Fig. 2; Abolafia-Rosenzweig et al. 2022a). The longitudes, latitudes, elevation, vegetation types, available observed spectrum bands of snow albedo (broadband, visible, and NIR), and atmospheric forcing variables for each site are provided in Table S1. Atmospheric forcing variables and surface albedo are hourly at three locations. Snow depth is available hourly at East River and Senator Beck sites, but it is daily at the Irwin site. The comprehensive methodologies for measuring solar radiation, albedo, and snow depth can be found in Abolafia-Rosenzweig et al. (2022a). At the Senator Beck site, manual snow profile measurements by the Center for Snow and Avalanche Studies (Landry et al. 2014; https://snowstudies.org/archival-data/) are used for model evaluation of snow water equivalent.
Locations of the three study sites with topography.
Citation: Journal of Hydrometeorology 26, 2; 10.1175/JHM-D-24-0082.1
The albedo and snow depth measurements have undergone rigorous quality control procedures to ensure their accuracy and reliability. To eliminate the negative effect of low sun angle on albedo, we use averages of albedo and snow depth measured between 1100 and 1300 local time (with the highest sun angle) for the three sites during the periods of investigation. The analysis is further limited to periods when the observed snow depth is more than 0.5 m at the East River site and 0.2 m in the Senator Beck and Irwin sites, to ensure that understory vegetation is completely buried by snowpack to eliminate the influence of vegetation on albedo, following Abolafia-Rosenzweig et al. (2022a). Unrealistic observed albedo values that exceed 1.0 or fall below 0.0 are removed.
b. MODSCAG product
We use a daily 463-m MODIS snow-covered area and grain size (MODSCAG) product (Painter et al. 2009; Rittger et al. 2020) to evaluate and constrain modeled snow grain size. Based on spectral unmixing and physically based snow radiative transfer models that remove soil and vegetation portions of the observed pixel, MODSCAG provides snow grain size at roughly 1030 LST (local solar time). For a clear sky day, the mean absolute error (MAE) for snow grain size from MODSCAG compared to field measurements at a single site is 51 μm (Painter et al. 2009). For both clear and cloud sky days, the gap-filled MODSCAG has a root-mean-square error (RMSE) of 118 μm for snow grain size compared to observations at three sites in the western United States (Bair et al. 2019). The MODSCAG data are obtained from the University of California, Santa Barbara (UCSB), website (https://snow.ucsb.edu/products/MODSCAG/WUS/) (last access: 29 May 2024). The snow grain size values are extracted from the MODSCAG grid cells covering the three study sites in this work.
c. Model setup and simulations
Noah-MP simulations adopt model-physics settings from the options used in recent continental-scale convection-permitting WRF/Noah-MP climate simulations (He et al. 2019a; Liu et al. 2017; Rasmussen et al. 2023) that have reasonably captured key hydroclimate and land surface states over the continental United States, except for different snow albedo options tested in this study. The snow-related parameters within Noah-MP follow the values used in the latest public release of Noah-MP, version 5.0 (He et al. 2023b). Leaf area index (LAI) is characterized by vegetation type based on a 10-yr monthly climatology of MODIS products (Yang et al. 2011). The vegetation type for each research site is grassland, and the canopy height is set as documented at each specific site (Abolafia-Rosenzweig et al. 2022a). For each study site, Noah-MP is first spun up for 11–13 years to get the steady state as listed in Table S1, followed by model analysis for subsequent years (Abolafia-Rosenzweig et al. 2022a). The analysis periods, dependent on the availability of observations, are October 2018–August 2021 in Irwin, October 2011–20 in Senator Beck, and July 2017–November 2019 in East River.
1) Atmospheric forcing
The Noah-MP simulations utilize atmospheric forcing derived from a combination of two sources: the hourly forcing data obtained from the 1-km observation-constrained NOAA’s Analysis of Record for Calibration (AORC) dataset (National Weather Service 2021) which are then replaced with in situ observed data when available at each study site. The required input atmospheric forcing variables include surface air temperature, surface specific humidity, wind speed in eastward/northward direction, surface pressure, downward shortwave and longwave radiation, and precipitation. The atmospheric forcing variables observed at each location are listed in Table S1. To minimize simulation uncertainty due to downward direct/diffuse shortwave radiation in visible and NIR bands, we use both the observed total downward shortwave radiation and the observed fraction of direct/diffuse and visible/NIR radiation as forcing to drive the model when observational data are available. If measurements do not provide observed direct/diffuse fractions, diffuse and direct radiation fractions are assumed to be 30% and 70%, respectively, to partition total downward solar radiation (Cuntz et al. 2016; He et al. 2023b).
2) Aerosol deposition flux
All model simulations are driven by the hourly aerosol (BC, dust, and OC) wet and dry deposition fluxes from the MERRA-2 reanalysis (Randles et al. 2017). MERRA-2 provides the wet and dry deposition fluxes for hydrophobic and hydrophilic (aged) OC and BC as well as dust with five size bins at a spatial resolution of 0.625° × 0.5°. The aerosol deposition fluxes at three in situ locations are determined by spatial interpolation based on the nearest neighbor grids of MERRA-2 values. Additionally, the sizes of MERRA-2 dust aerosol are converted to those compatible with the NoahMP–SNICAR setup (Table S2). We note that the MERRA-2 aerosol deposition data may have uncertainty but are a reasonable choice due to the lack of aerosol deposition measurements (Huang et al. 2022).
3) Fresh snow grain size optimization
We optimize the calibratable minimum and maximum values of freshly fallen snow grain size in SNICAR by comparing the simulated snow grain size to the MODSCAG data (section 3b). We conduct analyses for both fresh snow and aged snow cases. We define the fresh snow case based on the following criteria: 1) a daily observed increment of snow depth exceeding 0.02 m following Abolafia-Rosenzweig et al. (2022a), 2) a precipitation amount surpassing 0.0 mm day−1 following Wang et al. (2020), and 3) the model simulation indicates a complete (100%) snow-cover fraction. The original values are 54.526 μm for the minimum and 204.526 μm for the maximum of freshly fallen snow grain size. In this study, the minimal value is optimized to 33.0 μm, which is determined by identifying the smallest snow grain size from the MODSCAG data across all study sites and study periods. To optimize the maximum parameter value, we first assess the sensitivity of this parameter by conducting two simulations: i) increasing the default value by 50% and ii) reducing the default value by 50% to refine the parameter range. Subsequently, we use the “trial and error” method to adjust this parameter to 91.0 μm, aligning the modeled average fresh snow grain size with the MODSCAG averaged value over all study sites and study periods.
4) Model experiments
We conduct five model experiments (Table 1). The first simulation (Exp1) serves as the baseline case using the SNICAR snow albedo scheme. The SNICAR configuration includes the utilization of an adding-doubling solver; a midlatitude winter atmosphere profile with five spectral bands (300–700, 700–1000, 1000–1200, 1200–1500, and 1500–5000 nm); a hexagonal snow grain shape; the Colorado dust type; ice optical properties obtained from Picard et al. (2016); the inclusion of BC, dust, and OC; and the internal mixing of BC and dust with snow grains. Additionally, the simulation incorporates optimized parameters with respect to the freshly fallen snow grain size mentioned in section 3c(3). Subsequently, we perform four sensitivity simulations (Exp2–Exp5) to comprehend the impacts of snow grain size, snow grain shapes, and LAPs on snow albedo and surface radiative balance. The second simulation (Exp2) is identical to Exp1, except that it employs the original SNICAR parameters for freshly fallen snow grain size. The third simulation (Exp3) is identical to Exp1, except for changing the hexagonal snow grain shape to the spherical shape. The fourth simulation (Exp4) is identical to Exp1, except that it does not account for the influence of LAPs in snow. The final simulation (Exp5) utilizes the default Noah-MP BATS snow albedo scheme instead of SNICAR. All simulations are forced with the in situ observed snow depth to reduce the albedo uncertainty introduced by snow depth bias, as previous studies have shown nontrivial snow depth bias simulated by Noah-MP (Abolafia-Rosenzweig et al. 2021; Chen et al. 2014; He et al. 2019a, 2021; Ikeda et al. 2021). Specifically, the observed snow depth is directly inserted to ensure the simulated snow depth aligns with the observed values during periods when observations are available (Fig. S1). Previous studies (Abolafia-Rosenzweig et al. 2024; Hedrick et al. 2018; Wever et al. 2015) show that this strategy can enhance model performance for estimating snow depth and snow water equivalent. The snow depth is subsequently transformed into snow water and ice by multiplying with the modeled snow density, as typically done in snow depth data assimilation procedures. We note that this may introduce uncertainty to snow water equivalent in the model due to the lack of direct observations of snow density.
Noah-MP model configurations for different experiments.
d. Evaluation metrics
Statistical metrics are computed to assess the model performance. Bias is employed as a metric to assess the extent to which the modeled albedo is capable of properly replicating the average condition of observations. The RMSE is employed as a metric to assess the accuracy of the model capability in estimating the observed value. Pearson’s correlation coefficient r is used to quantify the consistency in temporal variability between modeled and observed snow albedo.
4. Results
a. Model evaluation
1) Snow grain size
The NoahMP–SNICAR simulation using original SNICAR grain size parameters (i.e., Exp2) produces systematically higher fresh and aged snow grain sizes by about 2 times compared to MODSCAG (Fig. 3). This discrepancy can be attributed to the overestimate of fresh snow grain size. When optimizing the size parameters of the freshly fallen snow grains [i.e., Exp1; see section 3c(3)], the modeled average fresh snow grain sizes (85 μm) at three research sites agree very well with the values (85 μm) obtained from MODSCAG (Fig. 3). Additionally, our results demonstrate that by improving the fresh snow grain size simulation in NoahMP–SNICAR, we can further effectively decrease the bias in the simulated aged (nonfresh) snow grain size. The mean value of the modeled aged snow grain size decreases from 182 to 103 μm, which matches very well with the observed value of 101 μm (Fig. 3).
Comparison of the average fresh and aged snow grain size obtained from MODSCAG and the simulated results from NoahMP–SNICAR simulations with the original and optimized fresh snow grain size parameters at three study sites during the periods when MODSCAG data are accessible. The uncertainty bar represents the spatiotemporal variability of snow grain size within one standard deviation. The fresh snow is defined in section 3c(3).
Citation: Journal of Hydrometeorology 26, 2; 10.1175/JHM-D-24-0082.1
2) Snow albedo during all periods
Overall, the NoahMP–SNICAR baseline simulation (i.e., Exp1 in Table 1) captures the observed mean snow albedo values, with a slightly higher broadband albedo by about 0.07 and less temporal variability (Figs. 4a–c and Table S3). The average bias is greater at the East River site compared to the other two sites. The broadband snow albedo overestimates mainly arise from the overestimated visible snow albedo by about 0.08, likely caused by the uncertainty in aerosol deposition and/or snow density, because snow grain size and snow depth are constrained by observations. This also explains the relatively good agreement between modeled and observed mean NIR snow albedo with the mean bias around 0.03 (Figs. 4a–c), since NIR snow albedo is sensitive to grain size. Uncertainty in snow grain shape could also slightly (by up to ∼0.04) contribute to the overestimated visible snow albedo based on sensitivity analysis [section 4b(2)]. The missing treatment of small-scale snow surface roughness in the model could also contribute to the snow albedo overestimates (Manninen et al. 2021), but it generally has a stronger impact on NIR albedo and hence may not be the main culprit here. Nevertheless, the NoahMP–SNICAR model accurately reproduces the observed phenomenon of the visible albedo being greater than the NIR albedo. The underestimated temporal variability of snow albedo at both visible and NIR bands is partially caused by the underestimated variability of snow grain size (Fig. 3), particularly during ablation periods (Figs. 4g–i). This is mainly due to the uncertainty in snow aging processes, which are less constrained by observations in this study. The uncertainty in aerosol deposition and evolution in snow could also contribute to the underestimated visible albedo temporal variability because the visible snow albedo is generally more sensitive to snow impurity than to snow grain size [section 4b(3)].
Site-level comparisons of (a)–(c) snow albedo, (d)–(f) fresh snow albedo, and (g)–(i) snow albedo over the melt period (Fig. S2) from observations (teal) and NoahMP–SNICAR (orange) and NoahMP–BATS simulations (green) at the (left) Senator Beck, (middle) Irwin, and (right) East River stations. The boxes are the interquartile ranges, the horizontal lines plotted in the boxes are the median values, and the whiskers indicate the maximum and minimum values of the results. Note that the green boxplots are very thin due to negligible temporal variability of the NoahMP–BATS snow albedo.
Citation: Journal of Hydrometeorology 26, 2; 10.1175/JHM-D-24-0082.1
3) Fresh snow albedo
The NoahMP–SNICAR baseline simulation of broadband fresh snow albedo reproduces the mean and temporal variability of observations due to the well-captured fresh snow grain size (Fig. 3), with higher accuracy in the Senator Beck and Irwin sites than in the East River site (Figs. 4d–f and Table S3). The simulated median broadband value closely matches the value of observed fresh-snow albedo at the Senator Beck site (0.88 observed vs 0.87 modeled) and the Irwin site (0.84 observed vs 0.82 modeled). The model also accurately simulates the interquartile range of the observed albedo values at these two sites. However, at the East River site, the modeled median value (0.83) is higher than the observed values (0.76) with underestimated temporal variability. For the visible band, the median fresh snow albedo is slightly overestimated by 0.03 at the Senator Beck and Irwin sites (Figs. 4d,e and Table S3). For the NIR band, the median fresh snow albedo is underestimated by about 0.02 at the Senator Beck site and about 0.05 at the Irwin site. Thus, the very small broadband biases at the Irwin and Senator Beck sites are attributable to the compensating errors in visible and NIR bands.
4) Snow albedo during melting periods
We evaluate snow albedo during the melting period, which is delineated as the time period from the peak of the observed snow depth (Fig. S2) to the end of July in each water year (1 October–30 September). As snow melts, the observed albedo decreases with increased temporal variability in comparison to fresh snow albedo (Figs. 4g–i). The simulated snow albedo generally captures the observations during melting periods at broadband, visible, and NIR bands, with a similar bias pattern as that of the entire snow period at the Senator Beck and Irwin sites (Figs. 4a,b) but a reduced bias at the East River site (Fig. 4c). Overall, the mean overestimated broadband albedo across three sites (by 0.1) is dominated by the overestimate in the visible band (mean bias = 0.12), with NIR albedo being better simulated (mean bias = 0.07). This is likely due to the uncertainty in aerosol content in snow, snow density, and/or snow grain shape as discussed in sections 4a(2) and 5. The underestimated temporal variability of snow albedo at all wavelength bands can be explained by the uncertainty in snow aging processes and aerosol content in snow as discussed in section 4a(2).
5) Comparison with default NoahMP–BATS snow albedo scheme
We further compare the NoahMP–SNICAR simulation with the Noah-MP simulation using the default semiphysical BATS snow albedo scheme (NoahMP–BATS) that has been recently optimized by Abolafia-Rosenzweig et al. (2022a). Overall, the NoahMP–SNICAR results outperform those of NoahMP–BATS at all three sites (Figs. 4 and 5 and Figs. S3–S5). The SNICAR scheme outperforms the BATS scheme in terms of temporal and geographic variabilities, as seen by the interquartile range and maximum–minimum values in Fig. 4. In Fig. 5, the slope and the correlation r in scatterplots between the observations and model simulations of snow albedo at all wavelength bands are enhanced in NoahMP–SNICAR relative to NoahMP–BATS. We note that the underestimated variability in the NoahMP–BATS snow albedo suggests inadequate physical linkage and sensitivity between snow albedo evolution and environmental/snowpack conditions in the BATS scheme, which is substantially improved by the SNICAR scheme. In terms of mean bias and RMSE, these values are reduced for the NIR band but not always for the visible band and the broadband at the Senator Beck site (Figs. 5a–c). However, the SNICAR scheme significantly reduces the NIR albedo overestimate at the Irwin site, resulting in a 50% improvement in mean bias and RMSE in the broadband, while the bias in the visible band remains unchanged (Figs. 5d–f). In the East River site, the mean broadband albedo bias and RMSE are also reduced (Fig. 5g). NoahMP–SNICAR improves the issue of conditional bias existing in NoahMP–BATS, i.e., the tendency of underestimating high albedo values and overestimating low albedo values, resulting in a reduction of root-mean-square error from 0.116 to 0.103 and an improvement in correlation from 0.42 to 0.67 (Table S4). Specifically, the SNICAR scheme enhances the variability in the visible snow albedo (Fig. S4), which mitigates the overestimate of visible snow albedo in BATS. In the NoahMP–BATS simulation, the visible snow albedo is consistently around 0.9 (Fig. 4 and blue dots in Figs. 5b,e), which is not realistic. This is because the BATS scheme uses a constant parameter for fresh snow albedo (Abolafia-Rosenzweig et al. 2022a; Wang et al. 2020). In contrast, in the NoahMP–SNICAR simulation, the fresh snow albedo is dynamically dependent on environmental conditions such as changes in temperature, snow depth, snow grain size, and the concentrations of LAPs. Furthermore, the simulation of NIR snow albedo is significantly improved by NoahMP–SNICAR relative to NoahMP–BATS, leading to a notable increase in the variability and hence a large reduction in bias.
Scatterplots comparing observed, NoahMP–BATS (blue dots), and NoahMP–SNICAR (red dots) simulated ground snow albedo in (a),(d),(g) broadband, (b),(e) visible, and (c),(f) NIR wavelengths at the (top) Senator Beck, (middle) Irwin, and (bottom) East River stations.
Citation: Journal of Hydrometeorology 26, 2; 10.1175/JHM-D-24-0082.1
b. Effects of snow grain size, snow grain shape, and LAPs on albedo and radiative forcing
Here, we quantify the modeled snow albedo and absorbed solar radiation at the surface in response to key snow albedo factors in NoahMP–SNICAR simulations.
1) Snow grain size
The optimization of fresh snow grain size parameters [section 3c(3)] leads to a decrease in snow grain size, which better agrees with observations (Fig. 3) and in turn increases snow albedo (Figs. 6a–c). The broadband snow albedo at the Senator Beck, Irwin, and East River sites increases on average by 0.022, 0.011, and 0.021, respectively (Table S5a). The albedo changes induce surface radiative forcing (SRF) values of −14.2, −6.1, and −12.9 W m−2 (Figs. 7a–c and Table S5b). The changes in snow grain size have a more pronounced impact on the NIR band compared to the visible band, which is consistent with previous studies showing higher NIR snow albedo sensitivity to snow grain size (e.g., Flanner et al. 2021). As a result, there are greater fluctuations in the SRF in the NIR band, leading to a decrease in the absorbed broadband solar radiation. In addition, the albedo and SRF changes induced by snow grain size changes are more pronounced for fresh snow compared to those in the melting period, mostly due to alterations in the fresh snow grain size by the parameter optimization.
Changes in snow albedo due to changes in snow grain size from original fresh snow grain parameters to (a)–(c) optimized ones, (d)–(f) snow grain shape from sphere to hexagonal shape, and (g)–(i) LAPs from no LAPs to with LAPs in three stations, (left) Senator Beck, (middle) Irwin, and (right) East River. The error bars represent the 95% confidence interval. The color of the plots represents the data for the entire snow season (All), as well as the cases for fresh snow and the melting period (Fig. S2).
Citation: Journal of Hydrometeorology 26, 2; 10.1175/JHM-D-24-0082.1
As in Fig. 6, but for the induced SRF.
Citation: Journal of Hydrometeorology 26, 2; 10.1175/JHM-D-24-0082.1
2) Snow grain shape
In contrast to a spherical shape, a hexagonal grain shape exhibits a greater snow albedo (Figs. 6d–f) and a lower SRF (Figs. 7d–f). This is because nonspherical grains have a smaller asymmetry factor and weaker forward scattering compared to their spherical counterparts (Dang et al. 2016; He et al. 2017, 2018b), which is more representative of real-world conditions (Flanner et al. 2021; Hao et al. 2023; He et al. 2024). The broadband snow albedo in the Senator Beck, Irwin, and East River stations increases by an average of 0.037, 0.042, and 0.039, respectively (Table S5a), due to the use of the hexagonal shape. This increase in albedo results in surface solar radiation absorption changes of −26.9, −29.5, and −26.2 W m−2, respectively (Table S5b). During the melting period, the influence of the snow nonsphericity has a greater impact on snow albedo and SRF compared to the time when the snow is fresh, due to the larger snow grain size and shallower snowpack during melting periods (He et al. 2018b; He 2022). The rise in albedo and the decrease in SRF caused by the hexagonal shape are more pronounced in the NIR band, because of the high sensitivity of the NIR albedo to snow grain shape (Dang et al. 2016; Flanner et al. 2021).
3) Light-absorbing particles
Figures 6g–i and 7g–i display the changes in snow albedo and SRF caused by the inclusion of LAPs, respectively. Overall, including LAPs in snow causes a decrease in the broadband snow albedo by an average of −0.012 at the Senator Beck station (Table S5a). This reduction in albedo results in an SRF of 8.7 W m−2 (Table S5b). The Irwin and East River sites exhibit greater changes in snow albedo and SRF compared to the Senator Beck site, mostly due to the higher concentrations of LAPs present in these two locations. The LAP-induced SRF at the Irwin station is 22.2 W m−2, while in the Easter River station, it is 16.7 W m−2. These changes correspond to a decrease in the broadband snow albedo of −0.031 in the Irwin station and −0.024 in the Easter River station. The effects of LAPs are much more pronounced in the visible band than the NIR band, consistent with the literature (e.g., Warren and Wiscombe 1980; Flanner et al. 2007). The melting period exhibits greater LAP-induced changes in snow albedo and SRF compared to the fresh snow period, because of larger snow grain sizes and higher snowpack density during melting periods (Flanner et al. 2021; He 2022) as well as the enrichment of LAPs as snow melts (Niu et al. 2017). In addition, the higher downward solar radiation during melting periods also contributes to the higher SRF compared to winter.
5. Uncertainty discussions and future directions
Snow grain shape, size, and snow LAPs all contribute to the potential uncertainty in snow albedo calculations and solar radiation processes. In three study sites, the shape of snow grain has a considerable impact on snow albedo over the whole snow season. However, because no model process accounts for the dynamic evolution of snow grain morphology and no direct observational constraints, the model’s assumption of nonspherical shape throughout the period is uncertain (He et al. 2024). In reality, the shape of snow grains demonstrates geographical variation and temporal variability, which necessitates additional refinements (Hao et al. 2023). We further analyze snow shapes from snow profiles at the Senator Beck site, showing that snow shapes were typically nonspherical during the accumulation period but more spherical during the melting period while the current model only has prescribed snow shapes without temporal evolution. Future studies are needed to consider the evolution of snow shapes, which is likely to reduce the snow albedo bias in NIR wavelengths.
During the melting phase, the snow albedo biases and the effects of LAPs on snow albedo are stronger than those during the accumulation period. Nonetheless, the coarse resolution of the MERRA-2 aerosol deposition data is associated with uncertainty. We compare the MERRA-2 dust deposition data at the Senator Beck site with observed mass of dust loading data (Reynolds et al. 2020; https://www.codos.org/#mass-loading-data) at the nearby Swamp Angel Study Plot from the Colorado Dust-on-Snow program. The underestimation of simulated snow albedo in visible wavelengths is likely at least partially attributable to the underestimation of aerosol deposition input from dust-on-snow events. The MERRA-2 reanalysis dust deposition data utilized in the NoahMP–SNICAR simulations are consistently lower than the mass of dust loading data observed at the nearby Swamp Angel Study Plot (Table S6). The global coarse resolution of MERRA-2 data (0.5° latitude × 0.625° longitude) cannot resolve the subgrid variability and point-scale high values over the specific location partially due to the smoothing of complicated terrain in the mountain (Oaida et al. 2015). Future integration of observed dust data into the model simulation would help reduce the bias in snow albedo. The LAPs in the snowpack are also influenced by tunable model parameters such as snow aging scaling factor and interlayer meltwater scavenging efficiency factor, both of which affect the size evolution of snow grains and the concentrations of LAPs within the snowpack through positive feedback mechanisms (Qian et al. 2014). Because of the lack of direct observations, these model parameters are poorly constrained and warrant further exploration to reduce uncertainty in calculating the interactive effects of grain size and snow LAPs on snow albedo particularly during melting periods.
Furthermore, it is important to acknowledge that atmospheric forcing data are fundamental yet unavoidable sources of uncertainty. We strive to utilize in situ observed forcing data to the greatest extent possible to decrease the level of uncertainty. However, precipitation measurements are not available for all study sites; precipitation is not monitored at the Senator Beck site due to wind exposure at the alpine location, but it is obtained from the Swamp Angel Study Plot in the Senator Beck Basin catchment where wind redistribution of snow cover is negligible (Landry et al. 2014). In Noah-MP, certain snowpack physical processes, such as densification, still have uncertainties (e.g., He et al. 2019a, 2021), which may contribute to the bias in the estimation of snow albedo. To mitigate atmospheric forcing and model physics uncertainty in this work, we consider it necessary to use observed snow depth data to constrain model simulations. With observed snow depth as constraints, the model could be closer to reproducing the right result with the right mechanism (i.e., reducing compensatory errors).
While we effectively reduce the propagation of snow depth biases in model simulations by the direct insertion method, this approach could attenuate the positive snow albedo feedback, likely diminishing the variability of snow albedo in model simulations. We additionally compare snow albedo from open-loop simulations with those from direct insertion (Table S4). The RMSE increases from 0.103 (direct insertion) to 0.166 (open-loop) for broadband albedo in NoahMP–SNICAR runs, but the correlation coefficients are comparable at 0.67 and 0.69, respectively. For NoahMP–BATS results, the RMSE rises from 0.116 to 0.134, while the correlation coefficient increases from 0.42 to 0.72. In both NoahMP–SNICAR and NoahMP–BATS simulations, there is a positive snow albedo bias from direct insertion simulations and negative snow albedo biases in open-loop simulations. Discrepancies between direct insertion and open-loop simulations are more pronounced in visible wavelengths relative to NIR. The positive snow albedo feedback amplifies the relatively lower albedo from the open-loop simulation, further enhancing radiative absorption due to larger grain size and earlier exposure to darker substrates. Overall, the open-loop simulations reduce the magnitude of snow albedo while increasing its variability. NoahMP–SNICAR exhibits no enhancement relative to NoahMP–BATS when the observed snow depth is not utilized to constrain the model due to compensatory biases in the NoahMP–BATS open-loop simulation (i.e., overestimated albedo partially compensating for snowpack underestimation).
The simple direct insertion method can produce inconsistencies between modeled variables, leading to water and energy balance errors due to direct adjustments to snow depth. The improvements from constrained model simulations presented herein may also be partially affected by errors arising from other uncertain snow processes such as snow compaction and density of fresh snow. The direct insertion method implicitly assumes that errors come exclusively from the model while observations are perfect, ignoring the uncertainties of observations. To understand these uncertainties, snow water equivalent is evaluated with observed snow profiles. The analysis of snow water equivalent indicates that model simulations underestimate the observed values from snow profiles at the Senator Beck site, but the discrepancies among model simulations are comparatively minor (Fig. S6). Although the observed snow depth data from the automated station is used to constrain model simulations, the simulated snow depth is still lower than the snow profile measurements (Fig. S7), resulting in the underestimation of snow water equivalent. This discrepancy arises from distinct locations of snow profile measurements and the automatic snow depth tower. At the Senator Beck site, wind redistribution of snow and terrain roughness had resulted in substantial variations in snow water equivalent and snow depth between the automated instrument tower measurements and adjacent snow profile plots (Landry et al. 2014). Given that solar radiation is measured at the tower as well, it is more representative to use automated snow depth measurements from the same location in our simulations which also have long-term records (Skiles et al. 2012). Moreover, the underestimation of snow density by the snow compaction scheme in the Noah-MP model could also lead to an underestimation of snow water equivalent (Abolafia-Rosenzweig et al. 2024). Overall, results indicate that simulated differences in snow albedo propagate to minor differences in snow water equivalent, relative to point-scale snow water equivalent biases attributable to other model uncertainties (e.g., wind redistribution and snow density). Future studies may explore more advanced data assimilation techniques such as particle filter and ensemble Kalman filter with multivariate and multisensors assimilation to circumvent these limitations (Girotto et al. 2023; Kumar et al. 2022).
Looking beyond this study, we plan to include SNICAR as a new option for the snow albedo scheme in the next community release version of Noah-MP, evaluate NoahMP–SNICAR over a larger domain, and conduct regional simulations in a coupled land–atmosphere modeling system to assess the feedback induced by NoahMP–SNICAR, such as the western United States that experience burning or/and regular dust-on-snow events (e.g., Gleason et al. 2019; Skiles et al. 2015).
6. Conclusions
To determine whether the hyperspectral snow radiation transfer scheme can improve the snow albedo simulation in the land surface model, we implemented the state-of-the-art snow albedo model, the latest version of SNICAR, into the refactored Noah-MP LSM, version 5, and evaluated in detail using ground measurements at three Rocky Mountain observation sites. The coupled NoahMP–SNICAR model physically accounts for the aerosol–snow–radiation interaction, snow grain growth and aging, and effects of snow grain size and shape on snow albedo. However, the representation of snow shape is assumed to be constant over time, which is a potential topic for future model enhancement. The NoahMP–SNICAR simulation well reproduces the observed broadband, visible, and NIR snow albedo, although it slightly overestimates the visible and broadband snow albedo. The SNICAR scheme significantly improves the temporal variability of snow albedo (particularly in the NIR band) compared to the default semiphysical BATS snow albedo scheme in Noah-MP (correlation increase from 0.42 to 0.67). The remaining bias in NoahMP–SNICAR could be attributed to uncertainties in the deposition and evolution of snow impurities and snow aging processes as well as atmospheric forcing and other potential snowpack physics (e.g., densification), which requires further studies. The individual impacts of snow grain size, nonspherical snow grain shape, and snow impurity on snow albedo and surface radiative forcing have different signs and magnitudes. Overall, the average changes in the broadband snow albedo due to the observation-constraining of fresh snow grain size, the use of nonspherical snow shape, and including LAPs at the three stations are 0.018, 0.039, and −0.022, respectively. This study substantially enhances the physical representations of snow albedo processes and evolution in Noah-MP, which offers a stronger snow albedo modeling capability for future studies. Future efforts include investigating the climate effects of aerosols in snow via land–atmosphere interaction and snow albedo feedback in fully coupled meteorology–chemistry–snow models.
Acknowledgments.
The authors thank three reviewers and the editor for their helpful comments on improving the paper’s quality. The authors declare no conflicts of interest. T.-S. Lin, C. He, and R. Abolafia-Rosenzweig acknowledge the support of NOAA’s Weather Program Office’s Subseasonal-to-seasonal (S2S) Grant NA22OAR4590503; NOAA’s Climate Program Office’s Modeling, Analysis, Predictions, and Projections Program (MAPP) Grant NA20OAR4310421; and the U.S. Geological Survey (USGS) Water Mission Area’s Integrated Water Prediction Program Grant 140G0121F0357. T.-S. Lin would like to thank Dr. Huilin Huang for the helpful conversations about the use of MODSCAG data and Exec. Dir. Jeff Derry and Ella Bump of the Center for Snow and Avalanche Studies of Silverton, Colorado (https://snowstudies.org/). T.-S. Lin would like to acknowledge the high-performance computing support from Cheyenne (https://doi.org/10.5065/D6RX99HX) provided by NSF NCAR’s Computational and Information Systems Laboratory, sponsored by the National Science Foundation. NSF NCAR is sponsored by the National Science Foundation. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of the National Science Foundation.
Data availability statement.
Noah-MP model code updates are publicly available (https://doi.org/10.5281/zenodo.11406420; Lin and He 2024a): https://github.com/tslin2/hrldas_snicar.git. Simulation data used in this manuscript are available at https://doi.org/10.5281/zenodo.11406443 (Lin et al. 2024b). In situ observed albedo and snow depth from the three study sites are available at https://data.mendeley.com/datasets/5393ck97d9/3 (Abolafia-Rosenzweig et al. 2022b).
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. M. Skiles, F. Chen, and D. Gochis, 2022a: 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.
Abolafia-Rosenzweig, R., C. He, and D. Gochis, 2022b: Evaluation and optimization of snow albedo scheme in Noah-MP land surface model using in-situ spectral observations in the Colorado Rockies, V3. Mendeley Data, accessed 28 June 2023, https://doi.org/10.17632/5393ck97d9.3.
Abolafia-Rosenzweig, R., C. He, F. Chen, and M. Barlage, 2024: Evaluating and enhancing snow compaction process in the Noah-MP land surface model. J. Adv. Model. Earth Syst., 16, e2023MS003869, https://doi.org/10.1029/2023MS003869.
Bair, E. H., K. Rittger, S. M. Skiles, and J. Dozier, 2019: An examination of snow albedo estimates from MODIS and their impact on snow water equivalent reconstruction. Water Resour. Res., 55, 7826–7842, https://doi.org/10.1029/2019WR024810.
Barnett, T. P., J. C. Adam, and D. P. Lettenmaier, 2005: Potential impacts of a warming climate on water availability in snow-dominated regions. Nature, 438, 303–309, https://doi.org/10.1038/nature04141.
Brun, E., 1989: Investigation on wet-snow metamorphism in respect of liquid-water content. Ann. Glaciol., 13, 22–26, https://doi.org/10.3189/S0260305500007576.
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.
Colbeck, S. C., 1982: An overview of seasonal snow metamorphism. Rev. Geophys., 20, 45–61, https://doi.org/10.1029/rg020i001p00045.
Conway, H., A. Gades, and C. F. Raymond, 1996: Albedo of dirty snow during conditions of melt. Water Resour. Res., 32, 1713–1718, https://doi.org/10.1029/96wr00712.
Cosgrove, B., and Coauthors, 2024: NOAA’s National Water Model: Advancing operational hydrology through continental-scale modeling. J. Amer. Water Resour. Assoc., 60, 247–272, https://doi.org/10.1111/1752-1688.13184.
CTSM Development Team, 2022: ESCOMP/CTSM: NEON release: Some NEON updates fixing AG sites, update MOSART, small fixes (ctsm5.1.dev114). Zenodo, https://doi.org/10.5281/zenodo.7342803.
Cuntz, M., J. Mai, L. Samaniego, M. Clark, V. Wulfmeyer, O. Branch, S. Attinger, and S. Thober, 2016: The impact of standard and hard-coded parameters on the hydrologic fluxes in the Noah-MP land surface model. J. Geophys. Res. Atmos., 121, 10 676–10 700, https://doi.org/10.1002/2016JD025097.
Dang, C., Q. Fu, and S. G. Warren, 2016: Effect of snow grain shape on snow albedo. J. Atmos. Sci., 73, 3573–3583, https://doi.org/10.1175/JAS-D-15-0276.1.
Dang, C., C. S. Zender, and M. G. Flanner, 2019: Intercomparison and improvement of two-stream shortwave radiative transfer schemes in Earth system models for a unified treatment of cryospheric surfaces. Cryosphere, 13, 2325–2343, https://doi.org/10.5194/tc-13-2325-2019.
Flanner, M. G., and C. S. Zender, 2006: Linking snowpack microphysics and albedo evolution. J. Geophys. Res., 111, D12208, https://doi.org/10.1029/2005JD006834.
Flanner, M. G., C. S. Zender, J. T. Randerson, and P. J. Rasch, 2007: Present-day climate forcing and response from black carbon in snow. J. Geophys. Res., 112, D11202, https://doi.org/10.1029/2006JD008003.
Flanner, M. G., K. M. Shell, M. Barlage, D. K. Perovich, and M. A. Tschudi, 2011: Radiative forcing and albedo feedback from the Northern Hemisphere cryosphere between 1979 and 2008. Nat. Geosci., 4, 151–155, https://doi.org/10.1038/ngeo1062.
Flanner, M. G., and Coauthors, 2021: SNICAR-ADv3: A community tool for modeling spectral snow albedo. Geosci. Model Dev., 14, 7673–7704, https://doi.org/10.5194/gmd-14-7673-2021.
Girotto, M., K. N. Musselman, and R. L. H. Essery, 2023: Data assimilation of seasonal snow. Applications of Data Assimilation and Inverse Problems in the Earth Sciences, Cambridge University Press, 79–92, https://doi.org/10.1017/9781009180412.006.
Gleason, K. E., J. R. McConnell, M. M. Arienzo, N. Chellman, and W. M. Calvin, 2019: Four-fold increase in solar forcing on snow in western U.S. burned forests since 1999. Nat. Commun., 10, 2026, https://doi.org/10.1038/s41467-019-09935-y.
Hall, A., 2004: The role of surface albedo feedback in climate. J. Climate, 17, 1550–1568, https://doi.org/10.1175/1520-0442pi re3j (2004)017<1550:TROSAF>2.0.CO;2.
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., and M. Flanner, 2020: Snow albedo and radiative transfer: Theory, modeling, and parameterization. Springer Series in Light Scattering, Springer, 67–133.
He, C., Y. Takano, K.-N. Liou, P. Yang, Q. Li, and F. Chen, 2017: Impact of snow grain shape and black carbon–snow internal mixing on snow optical properties: Parameterizations for climate models. J. Climate, 30, 10 019–10 036, https://doi.org/10.1175/JCLI-D-17-0300.1.
He, C., M. G. Flanner, F. Chen, M. Barlage, K.-N. Liou, S. Kang, J. Ming, and Y. Qian, 2018a: 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., K.-N. Liou, Y. Takano, P. Yang, L. Qi, and F. Chen, 2018b: Impact of grain shape and multiple black carbon internal mixing on snow albedo: Parameterization and radiative effect analysis. J. Geophys. Res. Atmos., 123, 1253–1268, https://doi.org/10.1002/2017JD027752.
He, C., F. Chen, M. Barlage, C. Liu, A. Newman, W. Tang, K. Ikeda, and R. Rasmussen, 2019a: Can convection-permitting modeling provide decent precipitation for offline high-resolution snowpack simulations over mountains? J. Geophys. Res. Atmos., 124, 12 631–12 654, https://doi.org/10.1029/2019JD030823.
He, C., K.-N. Liou, Y. Takano, F. Chen, and M. Barlage, 2019b: Enhanced snow absorption and albedo reduction by dust-snow internal mixing: Modeling and parameterization. J. Adv. Model. Earth Syst., 11, 3755–3776, https://doi.org/10.1029/2019MS001737.
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.
He, C., and Coauthors, 2023a: Modernizing the open-source community Noah with multi-parameterization options (Noah-MP) land surface model (version 5.0) with enhanced modularity, interoperability, and applicability. Geosci. Model Dev., 16, 5131–5151, https://doi.org/10.5194/gmd-16-5131-2023.
He, C., and Coauthors, 2023b: The community Noah-MP land surface modeling system technical description version 5.0. NCAR Tech. Note NCAR/TN-575+STR, 285 pp., https://doi.org/10.5065/ew8g-yr95.
He, C., M. Flanner, D. M. Lawrence, and Y. Gu, 2024: New features and enhancements in Community Land Model (CLM5) Snow albedo modeling: Description, sensitivity, and evaluation. J. Adv. Model. Earth Syst., 16, e2023MS003861, https://doi.org/10.1029/2023MS003861.
Hedrick, A. R., and Coauthors, 2018: Direct insertion of NASA airborne snow observatory-derived snow depth time series into the iSnobal energy balance snow model. Water Resour. Res., 54, 8045–8063, https://doi.org/10.1029/2018WR023190.
Huang, H., Y. Qian, Y. Liu, C. He, J. Zheng, Z. Zhang, and A. Gkikas, 2022: Where does the dust deposited over the Sierra Nevada snow come from? Atmos. Chem. Phys., 22, 15 469–15 488, https://doi.org/10.5194/acp-22-15469-2022.
Ikeda, K., and Coauthors, 2021: Snowfall and snowpack in the Western U.S. as captured by convection permitting climate simulations: Current climate and pseudo global warming future climate. Climate Dyn., 57, 2191–2215, https://doi.org/10.1007/s00382-021-05805-w.
Kumar, S. V., R. H. Reichle, C. D. Peters-Lidard, R. D. Koster, X. Zhan, W. T. Crow, J. B. Eylander, and P. R. Houser, 2008: A land surface data assimilation framework using the land information system: Description and applications. Adv. Water Resour., 31, 1419–1432, https://doi.org/10.1016/j.advwatres.2008.01.013.
Kumar, S. V., and Coauthors, 2022: An agenda for land data assimilation priorities: Realizing the promise of terrestrial water, energy, and vegetation observations from space. J. Adv. Model. Earth Syst., 14, e2022MS003259, https://doi.org/10.1029/2022MS003259.
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.
Lawrence, D. M., and Coauthors, 2019: The Community Land Model version 5: Description of new features, benchmarking, and impact of forcing uncertainty. J. Adv. Model. Earth Syst., 11, 4245–4287, https://doi.org/10.1029/2018MS001583.
Lin, T.-S., and C. He, 2024a: Model code for the journal publication titled “Improved snow albedo evolution in Noah-MP land surface model coupled with a physical snowpack radiative transfer scheme”. Zenodo, accessed 1 June 2024, https://doi.org/10.5281/zenodo.11406420.
Lin, T.-S., C. He, R. Abolafia-Rosenzweig, F. Chen, W. Wang, M. Barlage, and D. Gochis, 2024b: Model dataset for the journal publication titled “Improved snow albedo evolution in Noah-MP land surface model coupled with a physical snowpack radiative transfer scheme”. Zenodo, accessed 1 June 2024, https://doi.org/10.5281/zenodo.11406443.
Liou, K. N., Y. Takano, C. He, P. Yang, L. R. Leung, Y. Gu, and W. L. Lee, 2014: Stochastic parameterization for light absorption by internally mixed BC/dust in snow grains for application to climate models. J. Geophys. Res. Atmos., 119, 7616–7632, https://doi.org/10.1002/2014JD021665.
Liu, C., and Coauthors, 2017: Continental-scale convection-permitting modeling of the current and future climate of North America. Climate Dyn., 49, 71–95, https://doi.org/10.1007/s00382-016-3327-9.
Liu, L., M. Menenti, and Y. Ma, 2022: 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., 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.
Manninen, T., and Coauthors, 2021: Effect of small-scale snow surface roughness on snow albedo and reflectance. Cryosphere, 15, 793–820, https://doi.org/10.5194/tc-15-793-2021.
National Weather Service, 2021: Analysis of record for calibration: Version 1.1 sources, methods, and verification. Accessed 17 March 2021, https://hydrology.nws.noaa.gov/aorc-historic/Documents/AORC-Version1.1-SourcesMethodsandVerifications.pdf.
Niu, G.-Y., and Coauthors, 2011: The community Noah land surface model with multiparameterization options (Noah-MP): 1. Model description and evaluation with local-scale measurements. J. Geophys. Res., 116, D12109, https://doi.org/10.1029/2010JD015139.
Niu, H., and Coauthors, 2017: Distribution of light-absorbing impurities in snow of glacier on Mt. Yulong, southeastern Tibetan Plateau. Atmos. Res., 197, 474–484, https://doi.org/10.1016/j.atmosres.2017.07.004.
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.
Oleson, K. W., and Coauthors, 2013: Technical description of version 4.5 of the Community Land Model (CLM). NCAR Tech. Note NCAR/TN-503+STR, 434 pp., https://doi.org/10.5065/d6rr1w7m.
Painter, T. H., K. Rittger, C. McKenzie, P. Slaughter, R. E. Davis, and J. Dozier, 2009: Retrieval of subpixel snow covered area, grain size, and albedo from MODIS. Remote Sens. Environ., 113, 868–879, https://doi.org/10.1016/j.rse.2009.01.001.
Peters-Lidard, C. D., and Coauthors, 2007: High-performance Earth system modeling with NASA/GSFC’s Land Information System. Innovations Syst. Software Eng., 3, 157–165, https://doi.org/10.1007/s11334-007-0028-x.
Picard, G., Q. Libois, and L. Arnaud, 2016: Refinement of the ice absorption spectrum in the visible using radiance profile measurements in Antarctic snow. Cryosphere, 10, 2655–2672, https://doi.org/10.5194/tc-10-2655-2016.
Powers, J. G., and Coauthors, 2017: The Weather Research and Forecasting Model: Overview, system efforts, and future directions. Bull. Amer. Meteor. Soc., 98, 1717–1737, https://doi.org/10.1175/BAMS-D-15-00308.1.
Qian, Y., H. Wang, R. Zhang, M. G. Flanner, and P. J. Rasch, 2014: A sensitivity study on modeling black carbon in snow and its radiative forcing over the Arctic and Northern China. Environ. Res. Lett., 9, 064001, https://doi.org/10.1088/1748-9326/9/6/064001.
Qian, Y., and Coauthors, 2015: Light-absorbing particles in snow and ice: Measurement and modeling of climatic and hydrological impact. Adv. Atmos. Sci., 32, 64–91, https://doi.org/10.1007/s00376-014-0010-0.
Randles, C. A., and Coauthors, 2017: The MERRA-2 aerosol reanalysis, 1980 onward. Part I: System description and data assimilation evaluation. J. Climate, 30, 6823–6850, https://doi.org/10.1175/JCLI-D-16-0609.1.
Rasmussen, R. M., and Coauthors, 2023: CONUS404: The NCAR–USGS 4-km long-term regional hydroclimate reanalysis over the CONUS. Bull. Amer. Meteor. Soc., 104, E1382–E1408, https://doi.org/10.1175/BAMS-D-21-0326.1.
Reynolds, R. L., and Coauthors, 2020: Dust deposited on snow cover in the San Juan Mountains, Colorado, 2011–2016: Compositional variability bearing on snow-melt effects. J. Geophys. Res. Atmos., 125, e2019JD032210, https://doi.org/10.1029/2019JD032210.
Rittger, K., M. S. Raleigh, J. Dozier, A. F. Hill, J. A. Lutz, and T. H. Painter, 2020: Canopy adjustment and improved cloud detection for remotely sensed snow cover mapping. Water Resour. Res., 56, e2019WR024914, https://doi.org/10.1029/2019WR024914.
Robledano, A., G. Picard, M. Dumont, F. Flin, L. Arnaud, and Q. Libois, 2023: Unraveling the optical shape of snow. Nat. Commun., 14, 3955, https://doi.org/10.1038/s41467-023-39671-3.
Shao, D., W. Xu, H. Li, J. Wang, and X. Hao, 2020: Modeling snow surface spectral reflectance in a land surface model targeting satellite remote sensing observations. Remote Sens., 12, 3101, https://doi.org/10.3390/rs12183101.
Skiles, S. M., and T. H. Painter, 2019: Toward understanding direct absorption and grain size feedbacks by dust radiative forcing in snow with coupled snow physical and radiative transfer modeling. Water Resour. Res., 55, 7362–7378, https://doi.org/10.1029/2018WR024573.
Skiles, S. M., T. H. Painter, J. S. Deems, A. C. Bryant, and C. C. Landry, 2012: Dust radiative forcing in snow of the Upper Colorado River Basin: 2. Interannual variability in radiative forcing and snowmelt rates. Water Resour. Res., 48, W07522, https://doi.org/10.1029/2012WR011986.
Skiles, S. M., T. H. Painter, J. Belnap, L. Holland, R. L. Reynolds, H. L. Goldstein, and J. Lin, 2015: Regional variability in dust-on-snow processes and impacts in the Upper Colorado River Basin. Hydrol. Processes, 29, 5397–5413, https://doi.org/10.1002/hyp.10569.
Skiles, S. M., M. Flanner, J. M. Cook, M. Dumont, and T. H. Painter, 2018: Radiative forcing by light-absorbing particles in snow. Nat. Climate Change, 8, 964–971, https://doi.org/10.1038/s41558-018-0296-5.
Sthapit, E., and Coauthors, 2022: Evaluation of snow and streamflows using Noah-MP and WRF-hydro models in Aroostook River Basin, Maine. Water, 14, 2145, https://doi.org/10.3390/w14142145.
Thackeray, C. W., and C. G. Fletcher, 2016: Snow albedo feedback: Current knowledge, importance, outstanding issues and future directions. Prog. Phys. Geogr., 40, 392–408, https://doi.org/10.1177/0309133315620999.
Toon, O. B., C. P. McKay, T. P. Ackerman, and K. Santhanam, 1989: Rapid calculation of radiative heating rates and photodissociation rates in inhomogeneous multiple scattering atmospheres. J. Geophys. Res., 94, 16 287–16 301, https://doi.org/10.1029/JD094iD13p16287.
Verseghy, D. L., 1991: Class—A Canadian land surface scheme for GCMS. I. Soil model. Int. J. Climatol., 11, 111–133, https://doi.org/10.1002/joc.3370110202.
von Kaenel, M., and S. A. Margulis, 2024: Evaluation of Noah-MP snow simulation across site conditions in the western United States. J. Hydrometeor., 25, 1389–1406, https://doi.org/10.1175/JHM-D-23-0211.1.
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.
Wang, W., C. He, J. Moore, G. Wang, and G.-Y. Niu, 2022: Physics-based narrowband optical parameters for snow albedo simulation in climate models. J. Adv. Model. Earth Syst., 14, e2020MS002431, https://doi.org/10.1029/2020MS002431.
Warren, S. G., 1982: Optical properties of snow. Rev. Geophys., 20, 67–89, https://doi.org/10.1029/RG020i001p00067.
Warren, S. G., 1984: Optical constants of ice from the ultraviolet to the microwave. Appl. Opt., 23, 1206–1225, https://doi.org/10.1364/AO.23.001206.
Warren, S. G., and W. J. Wiscombe, 1980: A model for the spectral albedo of snow. II: Snow containing atmospheric aerosols. J. Atmos. Sci., 37, 2734–2745, https://doi.org/10.1175/1520-0469(1980)037<2734:AMFTSA>2.0.CO;2.
Warren, S. G., and R. E. Brandt, 2008: Optical constants of ice from the ultraviolet to the microwave: A revised compilation. J. Geophys. Res., 113, D14220, https://doi.org/10.1029/2007JD009744.
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.
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.
Xiao, M., S. P. Mahanama, Y. Xue, F. Chen, and D. P. Lettenmaier, 2021: Modeling snow ablation over the mountains of the western United States: Patterns and controlling factors. J. Hydrometeor., 22, 297–311, https://doi.org/10.1175/JHM-D-19-0198.1.
Yang, Z.-L., R. E. Dickinson, A. Robock, and K. Y. Vinnikov, 1997: Validation of the snow submodel of the Biosphere–Atmosphere Transfer Scheme with Russian snow cover and meteorological observational data. J. Climate, 10, 353–373, https://doi.org/10.1175/1520-0442(1997)010<0353:VOTSSO>2.0.CO;2.
Yang, Z.-L., and Coauthors, 2011: The community Noah land surface model with multiparameterization options (Noah-MP): 2. Evaluation over global river basins. J. Geophys. Res., 116, D12110, https://doi.org/10.1029/2010JD015140.
Zhang, R., H. Wang, Q. Fu, P. J. Rasch, and X. Wang, 2019: Unraveling driving forces explaining significant reduction in satellite-inferred Arctic surface albedo since the 1980s. Proc. Natl. Acad. Sci. USA, 116, 23 947–23 953, https://doi.org/10.1073/pnas.1915258116.
