| Severe convective weather is one of the meteorological disasters. It happensduring the short duration and is very destructive. This disaster weather brings hugelosses to people’s lives and property, communications facilities, agriculturalproduction and air transport. Up to now, there is no effective way to artificiallyweaken and control it, thus taking full advantage of information is particularlyimportant to forecast strong convective weather.With the development of science and technology, various weather forecastingmethods have been emerged and weather monitoring tools provide more and moreinformation. How to use numerous methods and information to make more accurateforecasts is the focus of people’s attention.In this paper, digital image processing technology was used to extract24-dimensional features of severe convective clouds from the Doppler radarreflectivity image. The feature data was imported to the database. Then data miningtechnology was used to extract rules from the feature database. The rules wereexpressed in the form of knowledge base. According to knowledge base anduncertainty reasoning in artificial intelligence, identification and prediction of strongconvective weather were achieved.The major work included:(1)Radar base data was read and transformed into the bitmap. According tometeorological knowledge, digital image processing technology was used to extract24-dimensional features of severe convective clouds from the Doppler radarreflectivity image. Then ADO was used to save the feature data to build the radarfeature database based on Access.(2)Rough Set Theory and Association Rules were used to dig out useful rules thatcan form the comprehensive knowledge base. The knowledge base included the imagefeatures and physical characteristics that directly generated by the radar data.Objective forecast model based on Rough Set and objective forecast model based onAssociation Rules were constructed. It laid a basis for comprehensive forecasting.Meanwhile, decision table reduction was used and rules with a high value of decision-making were obtained.(3)The image features (including strength features, texture features and shapefeatures) and the physical characteristics that directly generated by the radar data(including VIL, VILD, strong echo center and strong echo center top) were integratedto do data mining analysis. The rules conflicts were resolved by the theory ofuncertainty reasoning. Thus comprehensive forecast can be made.To sum up, this paper made a tentative study on the comprehensive forecast ofsevere convective weather. And to some extent the results met the forecastingrequirements. It provided a new method for the identification and forecast of severeconvective weather. |
