Abstract
The prediction of occurrence of severe convective events allows to issue meteorological alerts in order to reduce potential catastrophic events. In many cases, numerical models for weather forecast are not able to efficiently simulate such events. Alternatively, considering the large amount and diversity of meteorological data, the employment of Data Mining techniques has become widespread. In the case of convective activity, the use of past data allows to identify characteristic patterns in the model forecasts by associating them to the corresponding field of density of occurrence of cloud-to-ground atmospheric electrical discharges. Such task is performed by a machine learning algorithm, in this case, a set of neural networks. As a consequence, these patterns can be detected in future forecasts generated by the numerical model, allowing to predict the occurrence of discharges, which are associated to convective activity. In the current work, the proposed approach was applied to BRAMS, a numerical model developed in the country for operational and research use in Meteorology. Consequently, the ability to predict the occurrence of discharges by the set of neural networks was analysed for some selected events as well as its usefulness as an ancillary tool in operational weather forecast.
Highlights
Atmospheric convective events may cause severe storms with very strong winds and large precipitation volumes causing floods or landslides, among other effects
Numerical models for weather forecast properly simulate the movement of atmospheric air masses, but are still limited in predicting convective events, being the quality of the forecast subjected to the experience of the meteorologist and to the availability of further meteorological data and images
Typical data mining difficulties were found like to deal with very large amounts of data, missing data, the need of high processing power in the training phase of the neural networks and the iterative process of adjusting preprocessing parameters using the results as feedback, repeating many times the cycle of data pre-processing, machine learning, test and analysis
Summary
Atmospheric convective events may cause severe storms with very strong winds and large precipitation volumes causing floods or landslides, among other effects. The huge volume of meteorological data and images that are generated by sensors onboard of satellites or at the ground, as in the case of automatic meteorological stations, makes their prompt analysis very difficult or even impossible by the meteorologist that intends to monitor in real time the state of the atmosphere or to predict it. This issue has been tackled with the application of several data mining techniques in order to scan and analyze such large amount of data, being nowadays a current topic of research in Meteorology. Typical data mining difficulties were found like to deal with very large amounts of data, missing data, the need of high processing power in the training phase of the neural networks and the iterative process of adjusting preprocessing parameters using the results as feedback, repeating many times the cycle of data pre-processing, machine learning, test and analysis
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: Revista Mundi Engenharia, Tecnologia e Gestão (ISSN: 2525-4782)
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
