How to accurately locate the storm surge intensity as well as the magnitude of the disasters it causes based on the existing data, and carry out accurate warning for possible storm surge in the future, so as to reduce the loss caused by the storm surge disaster, is a key technical link for coastal cities of all countries to establish the forecast and warning system of the storm surge. The international forecast of typhoon surge disaster is mainly based on the predicted value of the surge during the impact of typhoon and then predict the highest tide level. Then, the value of the tide level is compared with the local warning water level. If the highest tide level will reach or exceed the warning water level, it is believed that storm surge disaster will occur and storm surge disaster forecast will be issued. The traditional correlative calculation is based on Gumbel, Weibull, Pearson-III models to analyze the wave height series, and to calculate the once-in-many-year surge and tidal level. When different probability models are used to analyze, the wave height values are very different. This is because no matter what kind of model is used for analysis, it is inevitable to have human factors and priori, and the results of different probability models are quite different. According to these forecasting models, storm surge disaster early warning will inevitably differ from the actual situation. Based on the principle of maximum entropy, a Poisson-maximum entropy compound extreme value distribution model based on the fusion of Poisson distribution and maximum entropy is established, and the model is applied to the classification of typhoon surge disaster intensity in Qingdao area. The recurrence period of typhoon surge at an observatory in Qingdao from 1963 to 1987 is calculated. Compared with the results obtained by other methods, the results obtained by this method are similar to those obtained by other methods, but the results obtained by this method are stable in the case of short data years. The new model is simple in form and contains several undetermined parameters. It can not only fully respect the measured data, but also use the known observation data to determine the parameters. It can constrain the existing data while maintaining the greatest uncertainty for the unknown information. In this way, the priori and artificial nature of the traditional methods can be avoided to a certain extent, so as to make the prediction results more reasonable.

Compound extreme value distributions; Disaster levels; Extreme distribution; Forecasting models; Principle of Maximum Entropy; Probability models; Storm surge disasters; Storm surges

Ocean University of China