1.IntroductionSnow albedo describes how much incoming solar radiation is reflected at the snow surface. It is determined by microphysical processes in addition to the macroscopic structure of the snowpack. For example, the variation of snow grain radius and grain shape affect snow optics and the proportions of scattered and absorbed solar energy (Dang et al., 2015; He, Flanner, et al., 2018; Wiscombe & Warren, 1980). The light scattering capability of a snow grain particle is also dependent on the wavelength, λ, for example, ice is almost non-absorptive in the ultra violet and visible wavelengths, where the single scattering albedo (ω), that is, the ratio of scattering efficiency to extinction efficiency is close to 1, but drops nonlinearly at near infrared wavelengths.As snow ages, the snow grain size increases and the albedo decreases. Under different temperature gradients, the snow grain would evolve into a ball or a lamellar structure (Rasmus, 2005). This metamorphic process results in stratified snow layers. Light-absorbing particles (LAPs) such as black carbon and mineral dust are enriched in the AbstractAccurate snow albedo simulation is a prerequisite for climate models to produce reliable climate prediction. Climate models would benefit from schemes of snowpack radiative transfer that are responsive to changing atmospheric conditions. However, the uncertainties in the narrowband snow optical parameters used by these schemes have not been evaluated. Conventional methods typically compute these narrowband parameters as irradiance-weighted averages of the spectral snow optical parameters, with the single scattering albedo being additionally weighted by the optically thick snowpack albedo. We first evaluate the effectiveness of the conventional methods as adopted by the widely used Community Land Model (CLM). Snow albedo calculations using the CLM narrowband optical parameters are relatively accurate for very thin snow (e.g., a bias of 0.01 for a 2-cm snowpack). The error, however, becomes larger as snowpack thickens (with biases of up to 0.05 for semi-infinite snowpack), because the snow radiative transfer is highly nonlinear and is most significant at wavelengths <1 μm. In this study, we propose a novel method to retrieve broadband optical parameters according to snow radiative transfer theory, reducing the albedo biases to <0.003 for 2 cm snowpacks and <0.005 for thick snowpacks. We find little impact in changing incident spectra on narrowband snow albedo. These newly derived narrowband optical parameters improve snow albedo accuracy by a factor of 10, allowing to trace the impacts of aerosol pollution in snow. The parameters are independent of which two-stream approximation is used, and are thus applicable to sea ice, glaciers, and seasonal snow cover.Plain Language SummarySnow albedo describes how much sunlight is reflected at the snow surface, which depends on how deep the sunlight penetrates the snowpack. Radiative transfer schemes describe sunlight absorption with snow optical depth. Snow radiative transfer schemes used in climate models make approximations using narrow-band snow optical properties for computational efficiency. A conventional way to derive the narrowband parameters is to average the wavelength-dependent values weighted by the incident solar spectrum. This approach produces snow albedo biases of up to 0.01 for shallow snowpacks and biases of up to 0.05 for thick snow. Such precision is not accurate enough for resolving the strongly positive snow-climate feedback when albedo decreases due to light-absorbing particles. This can amount to 0.01 over some “hot spots,” which are climatically significant and have received increasing attention. Here, we provide a new set of narrowband optical parameters that improve the snow albedo accuracy by a factor of 10.WANG ET AL.© 2021 The Authors. Journal of Advances in Modeling Earth Systems published by Wiley Periodicals LLC on behalf of American Geophysical Union.This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivsLicense, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.Physics-Based Narrowband Optical Parameters for Snow Albedo Simulation in Climate ModelsWenli Wang1, Cenlin He2, John Moore3,4, Gongxue Wang5, and Guo-Yue Niu61Key Laboratory of Regional Climate-Environment Research for East Asia, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China, 2Research Applications Laboratory, National Center for Atmospheric Research, Boulder, CO, USA, 3College of Global Change and Earth System Science, Beijing Normal University, Beijing, China, 4Arctic Centre, University of Lapland, Rovaniemi, Finland, 5Institute of Geospatial Information, Information Engineering University, Zhengzhou, China, 6Department of Hydrology and Atmospheric Sciences, The University of Arizona, Tucson, AZ, USAKey Points:The semi-empirical method used by Community Land Model to calculate narrowband snow optical parameters can produce errors that grow with snow massThe albedo errors stem from the relatively small biases in narrowband optical parameters that are amplified by nonlinear radiative transferWe propose a new set of narrowband snow optical parameters based on snow radiative transfer theory to improve albedo calculation accuracyCorrespondence to:W. Wang,wangwl@tea.ac.cnCitation:Wang, W., He, C., Moore, J., Wang, G., & Niu, G.-Y. (2022). Physics-based narrowband optical parameters for snow albedo simulation in climate models. Journal of Advances in Modeling Earth Systems, 14, e2020MS002431. https://doi.org/10.1029/2020MS002431Received 4 DEC 2020Accepted 19 OCT 2021Author Contributions:Conceptualization: Wenli WangFormal analysis: Wenli WangInvestigation: Wenli WangMethodology: Wenli WangSupervision: Guo-Yue NiuValidation: Wenli WangVisualization: Gongxue WangWriting – original draft: Wenli WangWriting – review & editing: Cenlin He, John Moore, Guo-Yue Niu10.1029/2020MS002431RESEARCH ARTICLE1 of 21
Journal of Advances in Modeling Earth SystemsWANG ET AL.10.1029/2020MS0024312 of 21upper snow layers as the snowpack melts. Since the LAPs absorb visible light better than pure snow, even at con-centrations of a few parts per billion (Dang et al., 2017; Flanner et al., 2007, 2009; He, Liou, et al., 2018), changes in atmospheric loading and surface enrichment of LAPs are important for snow albedo variability over time.Snow albedo is also dependent on illumination conditions and incident solar spectra. The radiative transfer treat-ment of direct light differs from that of diffuse light, as the direct solar beam acts as an external heating source to the snowpack that decays exponentially with snow optical depth (Toon et al., 1989). The solar zenith angle determines the path of direct sunlight and affects albedo (Aoki et al., 2003). The snow albedo under clear skies at all wavelengths increases with solar zenith angle (Wiscombe & Warren, 1980). The snow albedo under clear skies is lower than under cloudy skies for solar zenith angles <49.5° (Dang et al., 2015). Clouds, water vapor, and aerosols in the atmosphere reflect and absorb solar energy at various wavelengths, altering the incident solar spectrum reaching the snow surface. The mix of direct light and diffuse light therefore requires a merged solution in the snow radiative transfer scheme.The positive snow-albedo feedback between reduced surface albedo and atmospheric warming (Déry & Brown, 2007; Fletcher et al., 2012; Hall & Qu, 2006) has accelerated Arctic sea ice melting and mountain glacier retreat (Hanesiak et al., 1999; Li et al., 2017; Ming et al., 2015). Physical properties of snow such as depth, den-sity, grain size and shape, as well as the deposition of LAPs will change as a result of expected increases in fu-ture air temperatures and downward longwave radiation, reduced solid precipitation, and increased rain-on-snow events (Rasmus, 2005). Explicit treatment of snow micro-optical physics in the snow simulation modules used within climate models is needed to produce reliable macroscale characteristics, for example, snow cover fraction (Rasmus, 2005). To resolve the impacts of changing climate and LAPs on snow albedo requires snow albedo schemes of high, and as we shall show, higher accuracy than presently available in Land Surface Models (LSMs).Snow radiative transfer requires calculation of: (a) snow-scattering properties, including single scattering albedo (ω), extinction coefficient (e), and asymmetry factor (g), which are independent of radiative transfer schemes (Bohren & Barkstrom, 1974) and (b) radiative transfer process (e.g., two-stream approximations), which solves for radiative fluxes through the snow layers for both direct and diffuse radiation.Wiscombe and Warren (1980) developed a method to calculate spectrally resolved albedo across the 0.3–5 μm waveband for a homogeneous layer of pure snow. The nonspherical snow grains were represented by a collection of spheres with the same volume-to-area ratio (Giddings & LaChapelle, 1961). Flanner and Zender (2005) ex-tended Wiscombe and Warren (1980) into a multilayer model (Snow, Ice, and Aerosol Radiative Model—SNIC-AR) by solving a tridiagonal matrix solution for the two-stream radiative transfer scheme (Toon et al., 1989). Eight species of LAP (Flanner et al., 2007) and four snow grain shapes (He, Flanner, et al., 2018) have also been incorporated within SNICAR.Different from the radiative transfer approach, empirical or statistical formulations have been implemented into snow models to describe the evolution of snow albedo with time (Anderson, 1976). Some schemes include the effects of snow grain radius (Dang et al., 2015; Wang et al., 2020), grain shape (He et al., 2017; Liou et al., 2014), snow depth (Amaral et al., 2017; Wang et al., 2020) or snow density (Amaral et al., 2017), and LAPs (Dang et al., 2015; He, Liou, et al., 2018). These empirical schemes are unreliable for use in different environments and time periods because they are parameterized using limited observational data. Hence, empirical or semi-em-pirical approaches introduce various biases compared with the more process-based snowpack radiative transfer models (Wang, Yang, et al., 2021, companion manuscript).A reduced complexity snow radiative transfer model is better suited for use in climate models to minimize com-putational cost. For example, instead of the 10 nm spectral resolution employed in SNICAR, the Community Land Model (CLM; Flanner et al., 2007) uses a version of SNICAR with five narrow wavebands (0.3–0.7, 0.7–1.0, 1–1.2, 1.2–1.5, and 1.5–5 μm). Over the five wavebands, CLM computes the optical parameters (discussed in detail in Section 2) of snow and LAP as irradiance-weighted averages of the spectral optical parameters (at 10 nm resolution), then additionally weighted-averaging the single scattering albedo with the albedo of optically thick snow (Flanner et al., 2007). The optical parameters over the five narrow wavebands used in CLM are rather effective for 2-cm-thick snow layers, producing an albedo bias within 0.01 relative to fine-resolution (10 nm) SNICAR. But the error rises with increasing snow mass, as this averaging method is semi-empirical (Flanner et al., 2007) and does not preserve the accuracy of the fine resolution (spectrally resolved) SNICAR in the result-ing albedo over the five narrow wavebands. As more LSMs begin to adopt snow radiative transfer schemes for