Study of VLF/LF propagation anomalies in atmosphere by a network of ground-based sensors: a WEB platform for data distribution, analysis and visualization

A web platform for the distribution and visualization of VLF/LF data collected by ground-based sensors is presented. The web platform is intended to be used by the scientific community working on the study on anomalies in the propagation of VLF/LF signals in the atmosphere. Users can freely access the platform upon registration and contribute to the analysis of VLF/LF signals. A forum section is designed for discussions and interactions among users about the analysis of VLF/LF signals. Users are allowed to upload the data collected by their VLF/LF sensors after certification of data quality, besides downloading VLF/LF data from the platform.

A Virtual Private Server (VPS) is used to protect again potential breaches and unauthorized access to data. Users can select the location of VLF/LF transmitters and receivers, as well as the time window, and visualize the time series of signal amplitude (and phase if available).

Users can analyze the VLF/LF signals in terms of their wavelet spectrum. In addition, a few algorithmic tools are provided for the automatic detection of anomalies in the time domain. In this work we focus on a new algorithm which has been implemented, based on the Knorr algorithm which is more suitable for the analysis of multidimensional datasets with respect to statistics-based algorithms which require the knowledge of data statistical distribution.

The Knorr algorithm works as follows: given a data sample O in the time series T, it classified as outlier if a fraction p of data samples in T have a distance from O greater than a threshold D. Three types of outliers are defined:

• Global Outlier: the whole dataset is analyzed; this approach is more suitable for the off-line analysis of VLF/LF data;
• Left Outlier: samples already transmitted are analyzed; this approach is useful to detect anomalies in near-real time, with respect to the observed “normal” behavior of the time series;
• Right Outlier: samples transmitted after the occurrence of the anomaly are analyzed; this approach requires an off-line analysis of VLF-LF data and it provides the information about anomalies that have no longer been observed, within a given time window.