| The existing observation data suffers from temporal and spatial homogeneity problem due to different observation instruments and methods used, and global climate models (GCMs) can’t represent small-scale atmosphere features well due to the spatial and temperal resolution problem, which make regional climate research difficult. It is a good choice to drive the regional model which is suitable for describing the small-scale feature with the result from GCMs to obtain high-resolution regional climate feature. Now dynamical downscaling has been used in temperature and precipitatian widely, but not sufficiently in sea wind, especailly over the South China Sea (SCS). Firstly, the ability of simulating regional climate for the Weather Research and Forcast (WRF) Model has been tested by the simulation over SCS for the year 2003. Then two driving approaches and three popular global datasets (GD) have been compared to get the high-quality GD as driving field and the more suitable approach to attain the 10 km high-resolution sea wind data by dynamical downscaling. The added value of dynamical downscaling and the effect of spectral nudging have be assessed. The main conclusions are as follows:1) Continuous long-term integration simulation will lead to systematic errors accumulate, so the simulation results deviate from the observed values. Regularly re-initialization method can effectively decrease the error accumulated during long-term integration, with the spatial correlation coefficient (SCC) and root mean square error (RMSE) of monthly wind direction and speed being optimized.2)The modeled result of dynamical downscaling of ERA-Interim is better than that of FNL and CFSR, mainly reflected in the maximum annual average SCC (0.909) of wind speed and the minimum annual average spatial RMSE (17.99°) of wind direction, improving by 8.3% compared with the worst performance CFSR. Meanwhile, ERA-Interim have performed best in the RMSE and correlation coefficient (CC) of the regional average wind speed and direction. Compared with the worst wind speed simulation forced by FNL, the simulation of ERA-Interim have shown an improvement of 8.5% in RMSE of wind speed and 9.2% in CC of wind speed. Compared with the worst wind direction simulation forced by CFSR, the simulation of ERA-Interim have shown an improvement of 13.8% in RMSE of wind direction and 5.8% in CC of wind direction.3)Downscaling can help mitigate the problem that the wind speed intensity in global data is too weak and amend the spatial distribution of wind speed, such as fake maximum centers or maximum centers existing in the observation but not portrayed, to a certain extent. However, it is worth noting that downscaling may make the wind speed increase to a level, higher than the observed occasionally. Further assessment say that downscaling adds value to strong wind, but lack of ability to portray week wind correctly.4)Spectral nudging in this experiment has little effect, as a result of insignificant improvement of the large-scale circulation field. The spectral nudging coefficients may be set too low, and the use of five-daily re-initialization method shortens each integration time, which may block the effect of spectral nudging.5)Downscaling simulation from 2001 to 2008 showed two maximum center of frequency distribution of the high wind over the SCS, which are located in the Luzon Strait and the southeast coast of Vietnam with the maximum of 125 days. When the area with the frequency value higher than 75 days in the nothern SCS showed an obvious declining trend from 2001 to 2004 and an obvious growth trend after 2004, connected with the 75-day area in the southern SCS until 2008. The distribution of gale days presents obvious seasonal characteristics, with the maximum and growing trend in winter, minimum and decreasing trend in summer during the eight years... |
PDF Full Text Request