Statistical detection of meteor showers using data from the Swedish infrasound network Ludwik Liszka Swedish Institute of Space Physics SLU Umea, Sweden
Swedish Infrasound Network Present stations: Kiruna Jamton Lycksele Uppsala
Search for small and medium meteor events – objectives: To search for meteor events below the infrasonic background level To establish a set of indicators which can be used to discriminate a single event To develop techniques for extraction of events from combined multi-station data
Meteor on in Morjärv, Northern Sweden
Morjärv Meteor
Angle-of-arrival = 208º Vp = 350 m/s
Morjärv Meteor
How to optimize the indicators? Principal Component Analysis (PCA) of multiple indicator data Indicators are distributions of variables: Angle-of-arrival Phase velocity Crosscorrelation Spectral slope
Principal component analysis Purpose: to find a direction where maximum variance may be found in multivariate data
PCA Direction of maximum variance
Angle-of-arrival Entropy = (Emax=0, Emin=-2.42)
Phase velocity Vp
Cross-correlation
Spectral slope
Parameters of the analysis 1 window = 128 data points (7.11 sec) The window is moved in 32p steps (overlap) Distributions are created for each sample of 50 positions of the window 16 samples / 30 minutes
Discrimination of meteor impacts Selection of data to find proper combination of indicators: Concorde vs. North Sea Meteor Another approach is to apply PCA without pre-selecting transformation coefficients
Comparison of indicators
Plot of component loadings
High cross-correlation counts
Meteor discrimination
Bavarian Meteor th principal component
Discriminant function based on 5 th PC
Small meteors – meteor showers Leonids –
Small meteors – meteor showers Leonids –
Combination of information from 2 stations (North Sea Meteor) Lycksele - Jamton
Combination of information from 2 stations (North Sea Meteor) Lycksele - Kiruna
Combination of information from 2 stations (Leonids ) Lycksele - Jamton
Conclusions PCA may be used to discriminate events with a specific signature, like meteor impacts The method may be applied to events below the noise level, for example, meteor showers