Uncertainty of Land Surface Model and Land Use Data on WRF Model Simulations over China

The land–atmosphere interaction has been considered one of the most important part for weather prediction and climate modeling. To evaluate the uncertainty coming from land surface models (LSMs) and land use (LU) data in WRF simulated climatology over China, we have conducted fifteen 10-year simulations from 1996 to 2005 with three LSMs (NOAH, CLM and RUC) and five LU data sets (MODIS, HYDE, HH, RF and CESM). Compared to the MODIS, the most major differences for HYDE, HH and RF include the reduction of the barren or sparsely vegetated area and the CESM map shows the largest arid and semi-arid area. Based on performance evaluation of WRF model, the uncertainties of LSMs and LU data are analyzed in a three-dimension aspect: the magnitudes of response, spatial and temporal patterns. The impact of LSM and LU data is statistically significant in some regions and the LSM effect is substantially higher than the LU data especially for precipitation. The temporal effect of combinations of LSM and LU data varied across regions. For temperature, we find that the effects of LSMs and LU data on the spatial pattern and magnitude are one order smaller than those on temporal pattern, and the uncertainties from LSMs and LU datasets are as the same order when considering the temporal and spatial patterns. The results also indicate that the uncertainty of LU data on precipitation is much smaller than that of LSMs on magnitude and spatial patterns. These findings reflect that the relative importance of LSMs and LU data in the WRF climate modeling largely depends on the specific LSM.