A Database Engineered System for Big Data Analytics on Tornado Climatology

TL;DR

This paper introduces a database-engineered system utilizing RNN models to improve tornado prediction accuracy by integrating diverse climatology data, aiding in timely warnings and sustainable urban planning.

Contribution

It presents a novel database system that combines heterogeneous data sources with RNN-based forecasting for enhanced tornado prediction.

Findings

Accurately predicts tornado lead-time, magnitude, and location.

Demonstrates advantages of integrated data and RNN models in tornado forecasting.

Highlights system's effectiveness for big data analytics in climatology.

Abstract

Recognizing the challenges with current tornado warning systems, we investigate alternative approaches. In particular, we present a database engi-neered system that integrates information from heterogeneous rich data sources, including climatology data for tornadoes and data just before a tornado warning. The system aids in predicting tornado occurrences by identifying the data points that form the basis of a tornado warning. Evaluation on US data highlights the advantages of using a classification forecasting recurrent neural network (RNN) model. The results highlight the effectiveness of our database engineered system for big data analytics on tornado climatology-especially, in accurately predict-ing tornado lead-time, magnitude, and location, contributing to the development of sustainable cities.