设为首页
加入收藏
舟山群岛冬半年灾害性大风的成因与预报
The Causes of Catastrophic Gales in Zhoushan Islands with Their Forecasting
-
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。Abstract: 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.
-
Key words:
- Zhoushan islands;
- gale;
- neural network
-
图 1 2001年1月28日08:00风速与垂直速度沿30°N垂直剖面图
(粗实线:风速, 单位:m/s; 细线:垂直速度, 单位:10-1 Pa/s)
Fig. 1 Wind speed and vertical velocity distribution along the 30°N at 08:00 on Jan 28, 2001
(thick solid line denotes wind speed, units:m/s; thin line denotes vertical velocity distribution, unit:10-1 Pa/s)
图 2 2005年4月8日20:00 500 hPa涡度平流 (等值线, 单位:10-10s-2) 和850 hPa温度平流 (阴影, 单位:10-5 ℃·s-1)
Fig. 2 500 hPa vorticity advection (isoline, unit:10-10s-2) and 850 hPa thermal advection (shaded area, unit:10-5 ℃·s-1) at 20:00 April 8, 2005
图 3 有冷空气大风 (8级以上) 与无大风样本500 hPa与850 hPa风切变模的t统计量
Fig. 3 The t statistics of the wind shear modulus from 500 hPa to 850 hPa between the samples which have cold air gales (above 8 Beaufort scale) and no gale
-
[1] 曹美兰, 项素清. "晴天暴"的物理成因及预报.气象, 2002, 28(5):22-26. http://www.cnki.com.cn/Article/CJFDTOTAL-QXXX200205004.htm [2] 张晓慧, 盛立芳, 张红岩.渤海秋末初冬一次强寒潮天气过程分析.海洋预报, 2004, 21(3):51-56. http://www.cnki.com.cn/Article/CJFDTOTAL-HYYB200403007.htm [3] 孙建明, 陈卫锋.一次动量下传大风过程分析及预报着眼点.浙江气象科技, 2001, 22(4):1-4. http://www.cnki.com.cn/Article/CJFDTOTAL-ZJQX200104000.htm [4] 胡春蕾.95.11.7晴天暴过程分析.浙江气象科技, 1997, 18 (1):9-12. http://www.cnki.com.cn/Article/CJFDTOTAL-ZJQX199701002.htm [5] 仪清菊, 丁一汇.黄、渤海气旋爆发性发展的个例分析.应用气象学报, 1996, 7(4):483-490. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=19960474&flag=1 [6] 林明智, 李修芳, 余鹤书.预报爆发性气旋的一个综合判据.应用气象学报, 1993, 4(1):112-116. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=19930121&flag=1 [7] 林良勋, 程正泉, 张兵, 等.完全预报方法在广东冬半年海面强风业务预报中的应用.应用气象学报, 2004, 15(4):485-490. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=20040459&flag=1 [8] 祝启桓, 张淑云, 顾强民.浙江省灾害性天气预报.北京:气象出版社, 1992. [9] 王耀生.人工智能、模式识别在气象领域应用的现状与展望.气象, 1994, 20(6):9-14. http://www.cnki.com.cn/Article/CJFDTOTAL-QXXX406.001.htm [10] 张承福.人工神经网络在天气预报中的应用研究.气象, 1994, 20(6):43-67. http://www.cnki.com.cn/Article/CJFDTOTAL-QXXX406.007.htm [11] 王繁强, 徐文金, 陈杰伦, 等. B-P算法在青海省降雨分区分级预报中的应用.高原气象, 1997, 16(1):105-111. http://www.cnki.com.cn/Article/CJFDTOTAL-GYQX701.013.htm [12] 汤子东, 郑世芳, 奚秀芬.BP人工神经元网络在春季降水量预报中的应用.气象, 1997, 23(8):34-37. http://www.cnki.com.cn/Article/CJFDTOTAL-QXXX708.007.htm [13] 白慧卿, 方宗义, 吴蓉璋.基于人工神经网络的GMS云图四类云系的识别.应用气象学报, 1998, 9(4):402-409. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=19980460&flag=1 [14] 王成刚, 吴宝俊, 朱官忠.BP网络在鲁西南地区西南涡降水量级预报中的应用试验.气象科学, 1999, 19(1):158-165. http://www.cnki.com.cn/Article/CJFDTOTAL-QXKX199902006.htm [15] 王雷, 黄培强.利用人工神经网络预报芜湖的雾.气象科学, 2001, 21(2):200-205. http://www.cnki.com.cn/Article/CJFDTOTAL-QXKX200102009.htm [16] 施丹平.人工神经网络方法在降水量级中期预报中的应用.气象, 2001, 27(6):40-42. http://www.cnki.com.cn/Article/CJFDTOTAL-QXXX200106007.htm -
下载:
计量
- 摘要浏览量: 3072
- HTML全文浏览量: 664
- PDF下载量: 1686
- 被引次数: 0
