Research Of Cloud Classification And Intensity Forecast Of Tropical Cyclone On North-west Pacific

On the basis of the BP neural network model, cloud classification came from data of Infrared channel1, Infrared channel2and Water vapor channel of VISSR on the FY-2C geostationary meteorological satellite during2005and2009. The result was used to analysis the structure and the intensity of tropical cyclone.According to statistical theory, the sample probability distribution was assumed to help us to remove some apparent unreasonable data such as outliers, and to understand cloud normal features better. It was found that the value of IR1-IR2was most sensitive to the amount of thin cirrus cloud in the all6selected features because of the feature of its ability to figure out the cirrus cloud. On the other hand, the error of the BP neural network model also mostly came from the contradiction of this feature. On this basis, we designed the nested BP artificial neural network model composed of two layers, which the first layer contained with five features of IR1, IR2, WV, IR1-WV and IR2-WV that were used to classify each pixel to one of four categories such as Clear, Mixed cloud, Thick cirrus cloud and Strong convective cloud, and the second with one feature of IR1-IR2that were used to classify Mixed cloud to Low-level cloud, Middle-level cloud or Thin cirrus cloud. This paper gave the following conclusion:(1) The thin cirrus and middle-low cloud result of the nested neural network was proved better than the original network.(2) From generation to death of cyclone, cloud type changed significantly. High level cloud grew when low level decreased. Strong convective cloud gathered to the center and increased in the center until the eye generated.(3) Temperature of cloud changed little and the variance of temperature reached a maximum at strong tropical storm. The type and temperature of the center of the cyclone changed similar to the whole area. (4) The correlation of the average temperature and intensity of cyclone is good. Some relationship changed sharply in different stage.(5) Stepwise regression prediction of tropical cyclone intensity with the introduction of TBB factors made a positive effect. However, certain factors could be improved.