improved signal detection at seismometer arrays neil d. selby
TRANSCRIPT
![Page 1: Improved Signal Detection at Seismometer Arrays Neil D. Selby](https://reader031.vdocuments.mx/reader031/viewer/2022022519/6217df9a54938a7bbe476aed/html5/thumbnails/1.jpg)
Improved Signal Detection at Seismometer Arrays
Neil D. Selby
UK National Data Centre, AWE Blacknest
Page 1
![Page 2: Improved Signal Detection at Seismometer Arrays Neil D. Selby](https://reader031.vdocuments.mx/reader031/viewer/2022022519/6217df9a54938a7bbe476aed/html5/thumbnails/2.jpg)
Introduction
• Ongoing project to improve signal detection at seismometer arrays with the generalised F detector(Selby 2008, 2011):
– Essentially weighted least squares inverse problem using prior information.
– Likelihood ratio test using F statistic to find probability of detection.
• Here I describe the use of prior information (i.e. the physics of the problem) to set weights andequalise detection thresholds.
• I show results of a test of 130 array-days (10 days at 13 small IMS arrays) of waveform data usinga frequency-domain implementation (Selby, 2011).
• Future plans and further development.
Page 2
![Page 3: Improved Signal Detection at Seismometer Arrays Neil D. Selby](https://reader031.vdocuments.mx/reader031/viewer/2022022519/6217df9a54938a7bbe476aed/html5/thumbnails/3.jpg)
IMS seismometer arrays
Page 3
![Page 4: Improved Signal Detection at Seismometer Arrays Neil D. Selby](https://reader031.vdocuments.mx/reader031/viewer/2022022519/6217df9a54938a7bbe476aed/html5/thumbnails/4.jpg)
Aims
• To make optimal use of International Monitoring System seismometer arrays.
• Use a common baseline for all arrays, frequency bands, slownesses etc.
• Avoid interactive tuning; detection thresholds are based on physics of the problem - the priorinformation given to the inverse problem.
• Improve proportion of reliable detections:
– Increase number of reliable detections,
– Decrease number of (statistically) poor detections.
• Consequently improve automatic bulletins (event lists) by improving input to global associationalgorithm.
• Consequently reduce burden on analysts.
Page 4
![Page 5: Improved Signal Detection at Seismometer Arrays Neil D. Selby](https://reader031.vdocuments.mx/reader031/viewer/2022022519/6217df9a54938a7bbe476aed/html5/thumbnails/5.jpg)
Prior information...
• Array aperture and signal wavelength.
• Number of channels in each array.
• Noise power spectrum at each array.
• Noise correlation between the sensors at each array.
Page 5
![Page 6: Improved Signal Detection at Seismometer Arrays Neil D. Selby](https://reader031.vdocuments.mx/reader031/viewer/2022022519/6217df9a54938a7bbe476aed/html5/thumbnails/6.jpg)
Array aperture and signal wavelength
Page 6
![Page 7: Improved Signal Detection at Seismometer Arrays Neil D. Selby](https://reader031.vdocuments.mx/reader031/viewer/2022022519/6217df9a54938a7bbe476aed/html5/thumbnails/7.jpg)
Number of channels
Page 7
![Page 8: Improved Signal Detection at Seismometer Arrays Neil D. Selby](https://reader031.vdocuments.mx/reader031/viewer/2022022519/6217df9a54938a7bbe476aed/html5/thumbnails/8.jpg)
Noise power spectrum
• Seismic noise generally has a red powerspectrum.
• Slope of power spectrum varies withfrequency.
• Noise power spectrum varies with time andfrom array to array.
• Shape of power spectrum influences falsealarm rate.
• Adaptively whiten the noise to reduce falsedetections.
Page 8
![Page 9: Improved Signal Detection at Seismometer Arrays Neil D. Selby](https://reader031.vdocuments.mx/reader031/viewer/2022022519/6217df9a54938a7bbe476aed/html5/thumbnails/9.jpg)
Noise correlation
0
1
2
3
4
5
freq
uenc
y (H
z)
0 1 2 3
Sensor sep. (km)
Data
0
1
2
3
4
5
freq
uenc
y (H
z)
0 1 2 3
Sensor sep. (km)
Model
-1.0 -0.5 0.0 0.5 1.0
Correlation
• For an ideal array design noise isuncorrelated between sensors in the array.
• Often not the case, especially at smallerarrays.
• Correlated noise increases the chance ofbogus detections at certain slownesses.
• Counter using a model of correlation basedon azimuthally isotropic Lg propagation.
Page 9
![Page 10: Improved Signal Detection at Seismometer Arrays Neil D. Selby](https://reader031.vdocuments.mx/reader031/viewer/2022022519/6217df9a54938a7bbe476aed/html5/thumbnails/10.jpg)
Results
• F detector makes 50%of detections made bycurrent IDC system.
• However, 10% more Fdetections areassociated with REBevents.
Page 10
![Page 11: Improved Signal Detection at Seismometer Arrays Neil D. Selby](https://reader031.vdocuments.mx/reader031/viewer/2022022519/6217df9a54938a7bbe476aed/html5/thumbnails/11.jpg)
Results vary differ at each array
Page 11
![Page 12: Improved Signal Detection at Seismometer Arrays Neil D. Selby](https://reader031.vdocuments.mx/reader031/viewer/2022022519/6217df9a54938a7bbe476aed/html5/thumbnails/12.jpg)
Detection comparison by slowness and azimuth
Page 12
![Page 13: Improved Signal Detection at Seismometer Arrays Neil D. Selby](https://reader031.vdocuments.mx/reader031/viewer/2022022519/6217df9a54938a7bbe476aed/html5/thumbnails/13.jpg)
Papers
• Selby, N.D., 2008. Application of a generalized F detector at a seismometer array. Bull. seism.Soc. Amer., 98, 2469–2481.
• Selby, N.D. 2011. Improved Teleseismic Signal Detection at Small Aperture Arrays. Bull. seism.Soc. Amer., 101, (August 2011).
Page 13
![Page 14: Improved Signal Detection at Seismometer Arrays Neil D. Selby](https://reader031.vdocuments.mx/reader031/viewer/2022022519/6217df9a54938a7bbe476aed/html5/thumbnails/14.jpg)
Current and future development
• Time domain method has been developed for use at larger arrays.
• Preliminary testing on infrasound arrays promising.
• Real time testing and input to global association algorithm being considered.
• Future developments of method:
– Include signal correlation into the model,
∗ useful for Lg ,
– SASCs and/or amplitude and time “pit corrections”,
– Information about attenuation - t*,
– Broadband,
– Single channel detection.
Page 14