Lightning Prediction Using Artificial Neural Networks and Electric Field Mill Data
Abstract
Electric Field Mills (EFMs) located in the region surrounding Cape Canaveral record the electrification of the atmosphere near them. Research studying how these sensors could improve lightning warnings has had mixed results. This paper used a Convolutional Recurrent Neural Network (CRNN) and data from 30 EFMs from May-July of 2012-2016. The mean was calculated for every 60 second period and 30 minutes of this summarized data was used to create a lightning prediction with a warning period of 15 minutes. This method achieved a True Positive Rate (TPR) of 77.6%, a False Positive Rate (FPR) of 8.3%, a False Discovery Rate (FDR) of 48.1%, and an Operational Utility Index (OUI) of 53.9% (Kehrer et al., 2006). This suggests that the EFM sensor array, when used as a means to measure the electrification of the entire region, is capable of effectively predicting lightning for a 5-mile radius near Cape Canaveral. Moreover, achieving a 53.9% on the OUI rivals the best methods currently used implying that incorporating EFMs into lightning forecasting may reduced the FPR and save millions of dollars in delay and cancellation costs.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 02, 2018
- Accession Number
- AD1055963
Entities
People
- Daniel E. Hill
Organizations
- Air Force Institute of Technology