Enhancement of land surface information and its impact on atmospheric modeling in the Heihe River Basin, Northwest China

With increasing computational resources, atmospheric/environmental models continue to run at finer-grid spacing that can resolve land surface characteristics, such as topography, land use/land cover, and soil texture. This paper assesses the improvement in land surface information data sets and its impact on atmospheric modeling. The study focuses on the Heihe River Basin (HRB) in northwestern China. Fine-scale, remotely sensed, and in situ land surface data in HRB are derived and compared with the global data sets used in most mesoscale models. The incorporation of these fine-scale land surface data, compared to those currently used in MM5, yields substantially improved HRB land surface data sets. HRB local and regional data sets and the global land data set are used in a nonhydrostatic mesoscale model (MM5) to investigate the influences of land surface uncertainty on meteorological modeling in the lower atmosphere. Main results suggest the following: (1) enhanced land data sets have a stronger impact on atmospheric water vapor fields in the lower boundary layer than other meteorological fields. Soil texture data greatly impacts the local precipitation simulation and landuse data improves the air temperature simulation in the lower atmosphere; (2) generally, the average land surface temperature biases are reduced using the enhanced land surface information, but the low bias in zones with higher elevation and high bias in zones with lower elevation still persist; (3) the wet bias over rugged terrain and dry bias in the simulated water vapor in the flat plains are both reduced. Area mean bias of simulated accumulated monthly precipitation is greatly reduced using the enhanced soil data. Convective available potential energy was larger in the HRB mountain regions using the default land data, while it was decreased using the enhanced ones; (4) analysis of the correlation coefficient between simulation bias and the geographic features shows that there are some patterns in the simulation bias distribution. Generally, larger bias still exists in the foothills of the mountains.