Research On The Characteristics And Predicting Methods Of Severe Congvective Weather In Beijing Tianjin And Hebei Area

The conventional meteorological observations, the Dopplor Radar data, FY-2C, the NCEP reanalysis data and the BJ-RUC model output are used in this paper to analyze the severe convective weather in Beijing-Tianjin-Hebei (BTH) area. First, temporal and spatial features of the three severe convective weather distributions, namely, thunderstorm gale, hail and short-time strong rainfall are studied as well as the concentration degree and period, and the climatic statistical characteristics by the method of the Climate Statistical Analysis. Second, The synoptic patterns of the severe convective weather situation is objectively classification based on the Self-Organizing Maps (SOM). Then the equations of the severe convective weather predicting models are conducted through the average probability method (PRO), Artificial Neuron Network(ANN) and Artificial Neuron Network based on SOM(ASOM) methods. The models forecast abilities are also tested within 48h. The predicting model based on synthetic multi-indicator superposition is established by using the Mesoscale Conceptual Models of the hail. The single station forecast model of hail based on the Mesoscale Conceptual Models is established by using synthetic multi-indicator superposition method. At last, a earlier identifying approach of hail cloud is studied through the statistical analysis of the radar echo characteristics of hail cloud using improved TITAN algorithm based on mathematical morphology. The effictive method on the short-term forecasting and nowcasting of hail has been established based on the numerical forecast products output and radar parameters. The results are shown as below:(1) The Statistic Characteristics of the severe convective weather in Beijing-Tianjin-Hebei area.The statistical analysis results indicate:Strong wind weather mainly occurs generally from early May to mid-June, the hail weather mainly occurs from mid-June to early July, short-time strong rainfall mainly occurs from mid-July to mid-August; the three kinds of the severe convective weathers above-mentioned have the obvious characteristics of diurnal variation that occurs mainly from afternoon to early evening; the spatial distribution of severe convective weather are closely related to the topographic conditions.The hail day numbers during last 30 years has decreased obviously and 2.0-2.5a periodic oscillations for interannual variations, and the sharp mutation is in 1993. The negative trend of hail is that reduction is larger in the northern area than the southern areas, and larger in the mountain regions than the plain regions. The westward movement of upper jet streams score(200hPa), uplift of the 0℃layer and thickness decreases between 0℃layer and-20℃layer are main reason of frequency reduction of hail in BTH area in the past 30 years.(2) Objective synoptic patterns classification of the severe eonvective weather.Synoptic pattern of the severe eonvective weather in BTH from May to Septemper is objective divided into four types using SOM method:Warm and Wet Shear(WWS), Cold Vortex(CV), Northwest Airflow(NA), West wind trough(WWT). Among them, the types of NA and CV have the highest probability of severe eonvective weather, more than 65%, the type of WWS comes second, followed by WWT minimum.(3) Improve the traditional method of Selecting Forecast Factors and Artificial Neural Netowrk(ANN).The spatial distribution of correlation between forecast factors and predictand are analyzed. Basd on the study, the combinatorial factors extracted from the main influence region have biggish correlation coefficient and therefore the method can eliminate the limitations of the single-station factor. The analysis results show that the correlation between the new selected predictors and predictand has improved, even up to 17%. Improvement of ANN network can not only automatically determine network structure and the learning parameters, but also effectively solve the problem of poor generalization ability and possibility of misleading to local minimun in learning process. A related test of the fitting sinusoid show that the improved algorithm with the standard function fitting curve is much closer to, and the average error is smaller, fit better than the traditional ANN algorithm.(4) Potential forecasting modeling technology of the severe convective weather.The forecast results from the application of the three forecast methods to the new sample indicate that the prediction of improved ANN method is the best, ASOM method is the second, followed by the PRO method of forecasting which results is the worst, and the prediction performance of three methods are much more better during the day time (08:00-20:00, Beijing Time) than during the night(20:00-08:00, Beijing Time)(5) Analysis and diagnosis studies of a severe hail event.Before hail occurrence, the CAPE increases rapidly, and the trend ofθse wind profile is arcuate Correspondingly, at the same time, the CIN decreases gradually,0～3km vertical shear(SHR) increases significantly, thus the lower atmosphere becomes slightly unstable. During the mature stage of supercell storm causing hail, bow echo is found as well as weak echo region in the lower levwl, strong hanging echo in the middle-upper level and three-body scattering spike(TBSS). Vertical liquid water(VIL) has significantly increased before the hail.(6) The surface basic elements forecasting performance evaluation of Beijing rapid update cycle of assimilation forecast system(B J-RUC).The evaluation results show that BJ-RUC system has good forecast perfomance, from the verfication scores to surface observations, the forecasting results of the 2m temperature,10m wind speed and hourly precipitation are higher, but results of the 2m relative humidity is lower. The average error gradually increases (the forecasting perfomance becomes worse) with the increasing of the prediction time, and the forecasting pefomance of BJ-RUC is better within 0-12h than 12-24h. At the same time, the system has a good forecast for the diurnal variation of these meteological elements except hourly precipitation. The forecating capability to the 2m temperature is superior over other elements.(7) Single Station short-term forecasting technology of hail.The average critical success index (CSI), which is concluded according to the forecast result of four synoptic patterns, is 15.8%. Because the sampling limit, the lower proportion of the hail samples of the total sample, resulting in false alarm rate (FAR) higher. Meanwhile, the K index,(T-TD)850, V300,θse950 are better eliminate empty factors of hail weather forecast.(8) Nowcasting and earlier identifying technology of hail.The earlier identifying model of hail cloud is used to forecast whether the hail will occur within 0-30min, the critical success index (CSI) of forecasting results is 82%. the prewarning model of hail is used to forecast whether make warning of hail now, the CSI is 90%. Compared with the early identifying and early warning methods based on the single-parameter statistical threshold of hail cloud, The earlier identifying and warning method based on selected indicators have betther forecast effect.