Surface roughness length and background albedo are refined in the Weather Research and Forecasting Model (WRF). Default WRF assigns a fixed value to the surface roughness length and background albedo for each land-use category on every grid cell through the land surface scheme with only one urban category in USGS and MODIS land classification map. Such operation implies that all the urban area would be assigned the same value for surface roughness length and background albedo regardless the heterogeneous morphological properties, leading to an insufficient representation outcome for the urban area, especially in the Pearl River Delta (PRD) region with compact high-rise buildings and small urban village existing at the same time. Even though WRF can be coupled with urban canopy...[ Read more ]

Surface roughness length and background albedo are refined in the Weather Research and Forecasting Model (WRF). Default WRF assigns a fixed value to the surface roughness length and background albedo for each land-use category on every grid cell through the land surface scheme with only one urban category in USGS and MODIS land classification map. Such operation implies that all the urban area would be assigned the same value for surface roughness length and background albedo regardless the heterogeneous morphological properties, leading to an insufficient representation outcome for the urban area, especially in the Pearl River Delta (PRD) region with compact high-rise buildings and small urban village existing at the same time. Even though WRF can be coupled with urban canopy model (urban WRF) to make up for the lack of consideration of urban effects on wind speed and temperature, such measure can only be used in research simulation but not in the operational run of WRF due to its high cost of time and high demand of the input parameters.

With the problems mentioned earlier, a method to improve the wind speed and temperature prediction accuracy is being conducted to accommodate better the operational usage of WRF without exceptional extra time costs or a large number of parameters prepared. We calculate a new set of surface roughness length and background albedo, respectively, to replace the default WRF data by leveraging the buildings' data from Baidu Map and the Local Climate Zones (LCZs) developed by the World Urban Database and Access Portal Tools (WUDAPT) in PRD region. By integrating the new input data, surface roughness length, and background albedo, in the land surface scheme in the WRF, more precise wind speed and temperature prediction results are founded after running the default, modified simulations by comparing them with observation data in January and July. Since there is one method similar to improving the wind speed prediction that has already been proposed by Pak Shing Yeung, we also replicate this method to compare it with ours. The final result shows that for wind speed, there is approximately 37%, 60%, and 24% improvement in terms of root mean square error, mean bias, and index of agreement, respectively, compared to default ones. For temperature, we analyze the results from the method proposed in this study with the default model results. It shows 13%, 69%, and 6% improvement in terms of root mean square error, mean bias, and index of agreement, respectively.