Data Mining Approaches to the Real-Time Monitoring and Early Warning of Convective Weather Using Lightning Data

Tracking and monitoring of convective events may require the analysis of a huge amount of data from sensors on the ground, such as weather radars, or on board of satellites, in addition to the forecasts of numerical models. Thunderstorms associated with severe convective events have a potential to cause strong winds, floods, and landslides, with serious environmental and socio-economic impacts. New approaches based on data mining have been proposed for countries like Brazil that lack a complete weather radar coverage, but have some ground-based lightning detector networks. Lightning data may help to visualize the current state of convective systems in near real-time or to estimate the amount of convective precipitation in a given area and period of time. Data mining algorithms can be trained using numerical model data and lightning data yielding specific data mining models, which can be used to predict the occurrence of convective activity from numerical model forecasts. These data mining models may help meteorologists to improve the accuracy of early warnings and forecastings, mainly in countries that lack a complete weather radar coverage.