摘要
利用舟山市1994—2003年的实测风资料分5类统计了舟山群岛冬半年大风的发生规律。对一次冷空气个例进行诊断分析结果表明:大风是强冷平流、高空急流、动量下传等共同作用的结果。对一次低气压引起的大风的分析认为涡度平流、温度平流、潜热释放对低压发展有重大贡献。根据大风的成因和预报经验选择有关物理量进行t统计分析,选择有异常表现的物理量作预报因子。最后用人工神经网络方法建立预报模型,并进行了试报,试报误差都在4.5 m/s以下。
The Zhoushan islands see strong gales all the year round, which bring great dangers to sea operations. So it is very important to study the rules with the causes of strong gales, and set up methods to forecast strong gales. First the occurrence rules of strong gales of five kinds by ten years data of Shengsi station in Zhoushan city and the weather charts are studied. The results show that the gales caused by cold airs from the central area are more than the others, and the numbers have annual variation. The cold airs from the eastern areas are a little less than that from the central areas. The numbers of strong gales caused by depression in the East China Sea are much more homogeneous in each year. Diagnosis of a case of gales caused by cold airs shows that the causes are the combination of strong cold advection, upper air jet stream and momentum downward transport. Studies about gales caused by depression show that the vorticity advection, thermal advection and latent heat release make great contribution to the development of the depression. Some physical elements are selected to calculate the t statistic according to the types of strong gales and the forecast experiences. Then forecasting factors are selected based on the abnormal physical elements when gales occur. Forecasting factors are calculated by data interpolated from real station data. Then these factors are used to establish forecast models. For forecasting, the numerical production of T213 is used to calculate forecasting factors. The forecasting models are established by artificial neural networks, which have two layers of pass. The first transfer function is tangent, and the second is linear transfer function. 70% of the samples are used to train a neural network, 15% of which are used to verify the network, and the other 15 % are used to test. Finally the best neural network is selected according to the results of imitation. The correlation coefficients between real gales and imitations are all above 0.80. Otherwise, absolute errors of the test samples are all less than 4 m/s. The forecasting models have been used since October 2004. Up to March 2005, all the absolute errors of the forecast samples are less than 4.5 m/s. So it can be concluded that this method has some capability to forecast the gales in Zhoushan sea area.
出处
《应用气象学报》
CSCD
北大核心
2007年第1期80-85,共6页
Journal of Applied Meteorological Science
基金
浙江省舟山市科委和浙江省气象局联合资助科研项目"海岛突发性大风研究"资助。
