Electrical thunderstorm nowcasting using lightning data mining

Abstract

This paper presents a study developed at SIMEPAR (Paraná state weather service) using lightning data for electrical thunderstorm nowcasting. Thunderstorm electrical data collected at SIMEPAR such as lightning location, time of occurrence, current intensity, polarity, etc, are stored in real-time in a relational database. As a first step of this study, Microsoft Business Intelligence bundled with SQL Server 2005 Beta was used to access some of these data and create lightning clusters representing electrical thunderstorms. The clusters were continuously monitored to predict electrical thunderstorm displacement and evolution. Work was undertaken to assess suitability and reliability to the process. Algorithm parameters fitting and cases studies are under development and further work will be done using Weka clustering classes. Once approved, the methodology will be integrated to SisRaios - a Java lightning data visualization, analysis and thunderstorm monitoring and forecasting tool (presented in this paper). It is expected that this computational tool enhanced by the data mining study will aid meteorologists and power companies to monitor electrical thunderstorms, supplying information for starting up maintenance teams, as well as providing a better thunderstorm warnings in general and improving SIMEPAR's nowcasting capabilities.