Improved parameterization of snow albedo in WRF + Noah . Part II : 1 Applicability to snow estimates for the Tibetan Plateau 2

Abstract. Snow albedo is important to the land surface energy balance and to the water cycle. During snowfall and subsequent snowmelt, snow albedo is usually parameterized as functions of snow related variables in land surface models. However, the default snow albedo scheme in the widely used Noah land surface model shows evident shortcomings in land-atmosphere interactions estimates during snow events on the Tibetan Plateau. Here, we demonstrate that our improved snow albedo scheme performs well after including snow depth as an additional factor. By coupling the WRF and Noah models, this study comprehensively evaluates the performance of the improved snow albedo scheme in simulating eight snow events on the Tibetan Plateau. The modeling results are compared with WRF run with the default Noah scheme and in situ observations. The improved snow albedo scheme significantly outperforms the default Noah scheme in relation to air temperature, albedo and sensible heat flux estimates, by alleviating cold bias estimates, albedo overestimates and sensible heat flux underestimates, respectively. This in turn contributes to more accurate reproductions of snow event evolution. The averaged RMSE relative reductions (and relative increase in correlation coefficients) for air temperature, albedo, sensible heat flux and snow depth reach 27 % (5 %), 32 % (69 %), 13 % (17 %) and 21 % (108 %) respectively. These results demonstrate the strong potential of our improved snow albedo parameterization scheme for snow event simulations on the Tibetan Plateau. Our study provides a theoretical reference for researchers committed to further improving the snow albedo parameterization scheme.



44
The surface albedo directly determines the proportion of incident solar radiation that is 45 absorbed by the surface, and is an important parameter in climate and land surface 46 models (LSMs) (Sellers et al., 1996). Small changes in surface albedo can affect the 47 energy balance in the land-atmosphere system, and can drive both local and global 48 climate change (Bloch, 1964). 49 Surface albedo changes dramatically during snowfall and snowmelt cycles. Much 50 research has been carried out to identify the factors that influence these changes, 51 including the effects of terrain shielding, altitude, sky conditions, vegetation, and snow 52 properties such as grain size, liquid water content, depth, and impurities (Warren and LSM considers snow cover and age, but ignores other snow related factors, such as 88 snow depth, that can drive dramatic changes in albedo (Ek et al., 2003). This makes it 89 inappropriate to use the Noah LSM to characterize changes in snow albedo that follow 90 from snowfall and melt processes in complex topographic areas. However, the Noah 91 LSM appears to be the most readily available snow albedo scheme for long term climate variables in the Noah albedo parameterization scheme (Liu, 2020). This approach is not 108 the same as assimilating satellite retrieved snow related products into the LSM, which 109 has also been shown to lead to improvements (Xu and  ground observations. The aim of this study is to explore the potential of our improved 130 snow albedo parameterization scheme to simulate snow events over the whole Tibetan 131 Plateau more accurately than can be done using the standard default scheme. We hope 132 that this study will also provide a useful reference for researchers working to develop 133 and improve this, and other albedo parameterization schemes.       Fritsch cumulus parameterization scheme for clouds. 196 We conducted numerical experiments to simulate snow event 1 (EXP1), event 2 (EXP2),    Albedo is a key factor in determining the net radiation received at the surface, which  experiments also increased the correlation coefficient between observed and modeled 312 air temperature, by 0.01-0.07, which represents an improvement of 1-9 % (Fig. 3, Table   313 3).

314
Compared with using the default Noah snow albedo scheme, using the improved for calculations made at 1 km resolution than at 5 km resolution, and air temperature 320 estimates are more accurate at 1 km resolution than at 5 km resolution, regardless of 321 which albedo scheme is implemented (Fig. 3, Table 3). Therefore, fine resolution (i.e.,  simulations of albedo at 5 km resolution (Table 2). which is also closer to the observed mean of 0.3, than the mean of 0.6 calculated from 358 WRF using the default scheme (Fig. 4a, 4b). In general, the accuracy of the WRF 359 estimates when the new scheme is used is closely related to the observed albedo.

360
Compared with the WRF estimates made using the default Noah scheme, the WRF 361 estimates made using the improved scheme greatly reduce the overestimation of albedo 362 when the observed values are below 0.6, but seem to increase the underestimation of 363 albedo when the observed values are higher than 0.6 (Fig. 4c).    albedo scheme with an improved scheme is more significant for areas in the higher 426 resolution d02 model domain than for areas in the coarser d01 model domain (Fig. 6a).

427
Using the improved albedo scheme in WRF increases the correlation coefficient   which is closer to observations for EXP7 and EXP8 (Fig. 7). In general, WRF estimates  Implementing the improved snow albedo scheme in place of the default scheme greatly 582 decreases the overestimation of albedo from snowmelt to snow free processes, but does 583 not remove the underestimation of albedo during snowfall. This means that the 584 improvements mainly come from snowmelt and snow free simulations, and model 585 performance during snowfall may be worse when the improved albedo scheme is used.

586
This suggests an opportunity to further investigate how albedo is characterized by snow 587 depth and age in the snow albedo parameterization scheme.  paper.