Surface Albedo Assimilation and Its Impact on Surface Radiation Budget in Beijing

Surface albedo is a crucial parameter in land surface radiation budget. As bias exists between themodel simulated and observed surface albedo, data assimilation is an important method to improve the simulation results. Moreover, surface albedo is associated with the wavelength of the sunlight. So, solar radiation partitioning is important to parameterize the surface albedo. In this paper, the moderate resolution imaging spectroradiometer(MODIS-) retrieved direct visible, direct near-infrared, diffuse visible, and diffuse near-infrared surface albedoswere assimilated into the integrated urban landmodel (IUM).)e solar radiation partitioningmethodwas introduced to parameterize the surface albedo. Based on the albedo data fromMODIS and the solar radiation partitioningmethod, the surface albedo data set for the Beijingmunicipal area was generated. Based on the surface albedo data set and the IUM, the impacts of the surface albedo on the surface radiation budget were discussed quantitatively. Surface albedo is inversely proportional to the net radiation. For urban areas, after assimilation, the annual average net radiation decreases about 5.6%. For cropland, grassland, and forest areas, after assimilation, the annual average net radiations increase about 20.2%, 24.3%, and 18.7%, respectively.

In this paper, the MODIS-retrieved direct visible, direct near-infrared, diffuse visible, and diffuse near-infrared surface albedos were assimilated into the integrated urban land model (IUM). e solar radiation partitioning method [17] was introduced to parameterize the surface albedo to consider the effect of the wavelength of the solar radiation. Based on the surface albedo data from MODIS and the solar radiation partitioning method, the surface albedo data set for the Beijing municipal area was generated. Based on the surface albedo data set and the IUM, the impacts of the surface albedo on surface radiation budget were discussed quantitatively.

Study Area.
e study area of this paper is the Beijing municipal area. e research region is located at 39.3°-41.1°N, 115.2°-117.6°E. e terrain tilts from northwest to southeast (Figure 1(a)). e dominate land cover category (LUC) in each grid is based on MODIS 1 km resolution data in 2017 (Figure 1(b)).
e LUCs were transferred from International Geosphere-Biosphere Programme (IGBP) classification to United States Geological Survey (USGS) classification using the relationship listed in Table 1.
e Chinese Academy of Sciences 325 m-high Meteorology and Environmental Observation Tower (325 m tower) was located in downtown Beijing, and the longitude and latitude of the 325 m tower are 116.37 E and 39.97 N, respectively (Figure 1(a)).

Data.
e 325 m tower data were used for the surface albedo data comparison. e tower is located in downtown Beijing, and the altitude of the foot of the tower is 49 m. e radiation fluxes including the upward and downward shortwave and longwave radiation are measured using the radiometer at the 47-meter height. e time ranges of the observed upward and downward shortwave radiation are from March to October 2018. e observed surface albedo is calculated as follows: where α is the surface albedo; S↑ is the upward shortwave radiation (W m −2 ); and S↓ is the downward shortwave radiation (W m −2 ). e albedos for direct visible, direct near-infrared, diffuse visible, and diffuse near-infrared solar radiations are from MODIS in 2015. e temporal and spatial resolution of the MODIS albedo data are 8d and 1 km, respectively. e MODIS albedo products were spliced and clipped firstly; then, they were interpolated linearly from 8d to 1d. e atmospheric forcing data used to drive the land surface model are originated from the Global Land Data Assimilation System (GLDAS) [36]. e initial field data are also originated from GLDAS. e temporal resolution of GLDAS is 3 h. e spatial resolution of GLDAS data is 0.25°, which was interpolated to 1 km by using a bilinear interpolation method.

Land Model.
We employed the integrated urban land model (IUM) [37] to study the impact of the surface albedo on radiation budget. IUM was developed based on the common land model (CoLM) [38]. IUM integrates the land surface models for urban and natural land surfaces.

Surface Albedo Parameterization.
e surface albedo for bare soil [15] in the IUM could be described as follows: where h θ is the solar elevation angle ( o ); θ is the volumetric soil moisture; α vis,soil and α nir,soil are the albedos for visible and near-infrared solar radiations, respectively; and α soil is the albedo for bare soil. e surface albedo for urban in the IUM could be described as follows:    Advances in Meteorology where α vis,urb and α nir,urb are the albedos for visible and nearinfrared solar radiations, respectively, and α urb is the albedo for urban. e surface albedo for vegetated surfaces is the same as that in the CoLM. A two-stream approximation method is used to parameterize it. e grid point surface albedo is calculated as follows: where α veg is the albedo for vegetation and f soil , f urb , and f veg are the fractional cover for soil, urban, and vegetation, respectively.
In this paper, the moderate resolution imaging spectroradiometer-(MODIS-) retrieved direct visible, direct near-infrared, diffuse visible, and diffuse near-infrared surface albedos were assimilated into the integrated urban land model (IUM). e solar radiation partitioning method was introduced to parameterize the surface albedo. e assimilation algorithm is as follows: where R dv , R fv , R dn , and R fn are direct visible, diffuse visible, direct near-infrared, and diffuse near-infrared solar radiations (W m −2 ), respectively; they are calculated based on Weiss and Norman [35] and corrected to consider the effect of cloud cover [17]. α vis,dir , α nir,dir , α vis,dif , and α nir,dif are albedos for direct visible, direct near-infrared, diffuse visible, and diffuse near-infrared       Advances in Meteorology solar radiations, respectively; they are retrieved from MODIS. e surface albedo data set for the Beijing municipal area was generated based on the assimilation algorithm and MODIS data. e spatial and temporal resolution of the surface albedo data are 1 km and 1d, respectively.

Results and Discussion
First, in order to validate the accuracy of the surface albedo simulations before and after assimilation, we compared them with the observed surface albedos in the 325 m tower site. Figures 2(a) and 2(b) are the monthly and seasonal average surface albedo simulations before and after assimilation and the observations. e biases between the monthly average surface albedo simulations before and after assimilation and the observations are listed in Table 2. In the whole assimilation period, after assimilation, the albedo simulations are larger than those before assimilation. After assimilation, the biases are decreased apparently for all the months and seasons compared with those before assimilation.
en, we compared the spatial distribution of the annual and seasonal average surface albedo simulations before ( Figure 3) and after (Figure 4) assimilation in the Beijing municipal area. e differences (after assimilation minus before assimilation) between the simulation results before and after assimilation are plotted in Figure 5. For urban areas, the surface albedo simulations after assimilation are usually larger than those before assimilation. However, for other LUCs, the surface albedo simulations after assimilation are usually smaller than those before assimilation.
en, we compared the annual and seasonal average surface albedo simulations before and after assimilation for different LUCs. Here, to simplify the comparison, we merged the LUCs of the research region into four main LUCs, that is, urban, cropland, forest, and grassland (Table 3). For most of the urban areas, the annual average surface albedo simulations after assimilation are larger than those before assimilation ( Figure 6). However, for cropland, grassland, and forest areas, the annual average surface albedo simulations after assimilation are smaller than those before assimilation ( Figure 6). Positive correlations exist for all the four LUCs, and the correlation coefficients between the simulations before and after assimilation for urban, cropland, grassland, and forest LUCs are 0.447, 0.182, 0.557, and 0.570, respectively.
We also compared the monthly average surface albedos for these four LUCs (Figure 7). For urban areas, the surface albedo simulations after assimilation are larger than those before assimilation for all the 12 months. However, for cropland, grassland, and forest areas, the surface albedo simulations after assimilation are smaller than those before assimilation for all the 12 months. For cropland, grassland, and forest areas, the surface albedo simulations before assimilation are relatively small in summer and large in winter; on the contrary, the surface albedo simulations after assimilation are relatively large in summer and small in winter. For cropland, grassland, and forest areas, the surface albedo simulations before and after assimilation are relatively large in November because of the snowfall and December because of the snow cover ( Figure 8).
Surface albedo is associated with the upward solar radiation, so it is a crucial parameter in surface radiation budget. Figure 9 is the monthly average net radiation simulations before and after assimilation for the four main LUCs. Figure 10 is the scatter plot of the annual average albedo and net radiation simulation differences before and after assimilation for the four main LUCs (after        Advances in Meteorology assimilation minus before assimilation). Albedo is inversely proportional to the net radiation. For urban areas, where the surface albedo simulations after assimilation are larger than those before assimilation, the net radiation simulations after assimilation are smaller than those before assimilation. However, for other LUCs, the net radiation simulations after assimilation are larger than those before assimilation. Net radiation is not only associated with the upward solar radiation but also with the upward longwave radiation. As the upward longwave radiation is dependent on the surface radiative temperature, we compared the monthly average surface radiative temperature for these four LUCs (Figure 11). Figure 12 is the scatter plot of the annual average albedo and upward longwave radiation simulation differences before and after assimilation for the four main LUCs (after assimilation minus before assimilation). Albedo is inversely proportional to the upward longwave radiation. As the solar albedo increases, the solar radiation received by the land decreases; as a result, the surface radiative temperature and the upward longwave radiation decrease too. e monthly average surface radiation budget simulations before and after assimilation for the four LUCs are shown in Figure 13. A quantitative analysis was performed for the impacts of the surface albedo on the surface radiation budget (Table 4). For urban areas, compared with that before assimilation, after assimilation, the annual average net radiation decreases about 5.6%. For cropland, grassland, and forest areas, compared with those before assimilation, after

Summary
In this paper, the MODIS-retrieved direct visible, direct near-infrared, diffuse visible, and diffuse near-infrared surface albedos were assimilated into the IUM. e solar radiation partitioning method was introduced to parameterize the surface albedo. Based on the albedo data from MODIS and the solar radiation partitioning method, the surface albedo data set for the Beijing municipal area was generated. e surface albedo data were validated in the 325 m tower which was located in downtown Beijing. e result indicates that, after assimilation, the simulation results of the surface albedo are improved apparently.
Surface albedo is a crucial parameter in the surface radiation budget. e results indicate that the surface albedo is inversely proportional to the net radiation. For urban areas, compared with that before assimilation, after assimilation, the annual average net radiation decreases about 5.6%. For cropland, grassland, and forest areas, compared with those before assimilation, after assimilation, the annual average net radiations increase about 20.2%, 24.3%, and 18.7%, respectively.
In the near future, the observed four components of the solar radiation should be used to validate the solar radiation partitioning methods. More observational sites should be used to validate the surface albedo data set. e generation algorithm of the surface albedo data set will be extended to a larger area such as the whole of China or the world. Rn before assim S ↑ before assim L ↑ before assim Rn a er assim S ↑ a er assim L ↑ a er assim Rn before assim S ↑ before assim L ↑ before assim Rn a er assim S ↑ a er assim L ↑ a er assim Rn before assim S ↑ before assim L ↑ before assim Rn a er assim S ↑ a er assim L ↑ a er assim Rn before assim S ↑ before assim L ↑ before assim Rn a er assim S ↑ a er assim L ↑ a er assim (d) Figure 13: Spatial average of the monthly average radiation budget simulations before and after assimilation for the four main LUCs: (a) urban; (b) cropland; (c) grassland; (d) forest.

Advances in Meteorology
Data Availability e observational data from the 325 m tower station are from the Institute of Atmospheric Physics, Chinese Academy of Sciences, and are available in the supplementary materials.