Initial Perturbation For Ensemble Forecasts On Squall System

Dynamical mechanisms differ between systems in different scales. Based on the WRF model, the initial perturbation schemes are examined by selecting a typical meso-scale process, the squall line in order to search for proper initial perturbation criteria. The main conclusions are as follows:(1) The timing series weighted method assuming that the error increases in correspondence to the time lag is unreasonable, shown by the analyses of lagged errors. Methods using root mean square error as a scaling factor are utilized in three squall line cases. The RMSE method can not only improve the spread of members, but also reduce the root mean square error of each variable. Analysis of precipitation shows that the simulation via the modified SLAF scheme improves greatly (especially the center of the precipitation).(2) The key to ensemble forecasts is the way of choosing perturbed variables. Experiments on the fields of wind, potential temperature, and water vapor, respectively and altogether, indicate that outcomes depend not simply on the number of perturbed variables, with different perturbed variables for different systems.(3) The type and the development of severe convective systems largely depend on the scale of the environment. Instability parameters express the environmental characteristics from various aspects:thermal instability and the energy parameters mainly reflect the thermal structure of atmospheric conditions and the potential associated with the likelihood and intensity of convection; dynamic instability parameters of the dynamic environment reflect in the atmosphere conduction to the occurrence of strong convection and environmental field affecting the type of system. Therefore, the squall line perturbation criteria based on the thermal and dynamic instability are feasible.(4) When correlation coefficient between potential temperature and CAPE is greater than 0.5 at the lower atmosphere (below850hPa), wind shear at 6km grows linearly in the boundary layer, with wind shear values greater than 2* 10-3s-1, and BRN number less than 20, perturbed potential temperature is preferred; on the contrary, when the correlation coefficient is less than 0.5, the boundary layer to the 6km altitude wind shear at the non-linear growth, and wind shear values<2* 10-3s-1, BRN number>20, with large wind speed variability, the effect of perturbation on the potential temperature is no longer obvious. According to the criteria,2 cases have been tested; found that the prediction results are better than randomly selecting perturbed variables forecasts, when analizing the error of the cumulative surface precipitation, or the spread of the members.