| Extreme rainfall brought by typhoon can cause great disasters in China. With the development of social economy, the vulnerability of society greatly increases when facing natural disasters. Accurate forecast for typhoon extreme precipitation are urgently needed. Nowadays, high resolution numerical weather prediction model begins to show advantages in typhoon rainfall forecast. However, there are still several scientific questions remained unanswered. In this study, a typical case is selected to study the potential of high resolution NWP model in predicting extreme precipitation events and the influence of uncertainty in the large-scale initial field.First of all, Percentage Method is used to investigate the threshold values for extreme precipitation brought by tropical cyclone(TC), based on data from 1981 to 2013 for TCs affecting China. The features of extreme rainfall brought by typhoon Fitow(2013) and Soudelor(2015) are studied using these threshold values. The return periods are also analyzed. It is found that the heavy rainfall brought by Fitow exceeds the extreme threshold for hourly, daily and process rainfall in Shanghai and northern Zhejiang. The heavy rainfall brought by typhoon Soudelor exceeds the extreme threshold for daily rainfall at the joint location of Zhejiang and Fujian provinces. There are several stations with precipitation at 10-year return period level, with a few at 100-year return period level.The heavy rainfall brought by typhoon Soudelor is then selected to be simulated with WRFV3.6 at a double-nested of 9km and 3km. It is found that the simulation reproduces the track and intensity of Soudelor very well, as well as the precipitation at the joint location of Zhejiang and Fujian provinces. As compared with the observation, the simulated distribution of precipitation is generally good and the simulated 6-hour precipitation is weaker. The simulated 24-hour precipitation is very close to the observation in intensity, with a displacement of 100-200 km in position.The Fractions Skill Score(FSS) is used to quantify the capability of the model in reproducing the heavy rainfall, together with sensitivity experiments for several key parameters, including the total area for calculation, size of the fraction, and the accumulated hours for precipitation. It is found that FSS decreases with the increase of the total area for calculation, but increases with the increase of fraction size. FSS increases with the increase of accumulated hours for precipitation, implying that the longer the accumulated hours, the better the prediction.Analyses on the water vapor flux and potential vorticity fields demonstrate that these two quantities correspondence with the distribution of precipitation very well. There is a high lag-correlation between the area mean precipitation and water vapor flux(or potential vorticity), implying that the extreme precipitation of Soudelor occurred in a favorable dynamical and thermo-dynamical background.On the basis of the above simulation, 20 more simulations are carried out with the uncertainty in large-scale initial field being considered. It is found that the fine disturbances in large-scale initial field result in significant impacts on the prediction of the intensity and location of extreme rainfall. The forecast uncertainty increases with the increase of precipitation intensity. The maximum predicted 24 h rainfall extreme is 755 mm and the minimum is 514 mm. The largest FSS for 24 h rainfall is 0.87, with the smallest 0.74. The lead/lag correlation analyses for all the ensemble members show that the water vapor flux is more significant and stable than the potential vorticity in indicating the occurrence of heavy rainfall. |
PDF Full Text Request