Infrasound Workshop San Diego, USA 2730 October 2003

1
Infrasound WorkshopSan Diego, USA27-30 October
2003

• Infrasound Processing at the IDC using DFX-PMCC
• N. Brachet
• International Data Centre
• Waveform Development Unit
• Preparatory Commission for the Comprehensive
• Nuclear-Test-Ban Treaty Organization
• Provisional Technical Secretariat
• Vienna International Centre
• P.O. Box 1200, A-1400 Vienna, Austria
• E-mail Nicolas.Brachet_at_ctbto.org

2

Outline of the Talk

• Software design and signal review
• Started with the increased amount of infrasound
  data ( IMS stations processed at the IDC)
• Understand the signal characteristics with the
  up-to-date interactive WinPMCC (DASE)
• Software development
• Upgrade the existing DFX-PMCC (as delivered in
  IDC Release 1) using WinPMCC as a reference
• Establish a baseline configuration
• Identify key parameters that impact on the
  DFX-PMCC detection capability
• Propose generic DFX-PMCC parameters for
  processing IMS arrays
• Software testing and validation
• Reprocess the known infrasound events archived in
  the IDC reference database
• Large scale testing in a real-time environment

3

Milestones of DFX-PMCC at the IDC

• A few dates
• June 1998 Installation of DFX-PMCC at the PIDC,
  Arlington
• April 2001 First IMS station (I26DE) processed
  at the IDC
• July 2001 Delivery of WinPMCC V1.0.4
• June 2002 DFX-PMCC run offline, results
  monitored on IDC infrasound intranet
• Summer 2002 IDC analysts test WinPMCC -
  Validation of the method
• February 2002 Plans for design of future
  infrasound processing at the IDC
• 2003 DFX-PMCC/WinPMCC comparison - Bug
  corrections
• Major update of DFX-PMCC source code (3D results)
• September 2003 Baseline configuration of
  DFX-PMCC - validation on reference events
• November 2003 (Plan) DFX-PMCC on development
  LAN (real-time testing)

4

DFX-PMCC Software Development

• Major modifications of the structure of the
  software
• Produce and store 3D pixels containing the
  derived data (az,sp,f,t) into files
• Store the results into IDC database using
• The usual DFX tables ARRIVAL, DETECTION, WFDISC
• additional tables PMCC_FEATURES, DERVDISC,
  PMCC_RECIPE
• Enhancements and corrections of existing modules
  
• Improve the azimuth calculation for PMCC families
• Use relative speed instead of absolute speed when
  grouping the pixels (detection of seismic phases)
• handle dead channels or met. data in the
  processing
• Improve the module in charge of closing/merging
  large families
• Increase the maintainability Clean unused source
  code (10 of total) and comment source code

5

DFX-PMCC Software Development

• Synchronize DFX-PMCC and WinPMCC results
• Bring DFX-PMCC at the same level of confidence as
  WinPMCC (the reference)

Long March rocket launch detected at I34MN, 15
Oct 2003
Azimuth
Speed
DFX-PMCC
WinPMCC
6

Baseline Configuration for DFX-PMCC

• Key factors impacting on DFX-PMCC detection
  capability
• Preprocessing DFX Data Quality Control (QC)
• Frequency resolution Selection of frequency
  bands and filters
• Azimuth resolution
• Selection of the best sub-network combination
• Control the number of events per station size
  of signals to be detected
• Accomplishments

DFX-PMCC in Jan2003
Baseline config.
DFX-PMCC in Sep2003
7

Baseline Configuration for DFX-PMCC DFX-Quality
Control (QC)
Objective Remove the HF bogus detections

• Action Reduce DFX-QC action
• DFX-QC generates unstable bogus PMCC detections
  at high frequency
• DFX-QC searches and masks poor quality samples -
  Waveforms are repaired (interpolation) when
  possible
• Two levels of Data QC basic (spikes, gaps,
  sequences) and extended (complex spikes,
  multi-channel analysis)

• Turn off the extended QC and use only a basic QC
  for gaps, sequences and larger spikes
• Results Close to 100 of bogus HF detections
  have disappeared

8

Baseline Configuration for DFX-PMCC DFX-Quality
Control

• Objective Reduce the sensitivity of DFX-PMCC to
  microbaroms
• Preliminary observation understand the signal
  characteristics
• Microbaroms produced by marine storms
• Narrow band signals 0.1 - 0.4 Hz
• Series of signal bursts that can be detected
  over long time periods (typically several hours)
• More variation in detection azimuths when the
  source is closer to the station

9

Baseline Configuration for DFX-PMCC Frequency
Bands

• Action Use lower weight in the processing of
  specific freq. bands
• Do not use a band-stop filters
• Take advantage of the PMCC method for clustering
  pixels into detection families
• Detection pixels are those which verify the
  consistency relation
• Pixels with similar attributes (az,sp,f,t) are
  grouped into families
• Families are turned into detections if they
  contain a minimum number of pixels

10

Baseline Configuration for DFX-PMCC Frequency
Bands

• Results
• Trap the microbaroms detections in a narrow
  frequency band.
• Post-processing can be applied in order to
  identify these detections and not to use them in
  subsequent stages of processing (event building)
• Remove the smallest sequences of microbaroms

• Same logic can be applied to any narrow band
  signal

11

Baseline Configuration for DFX-PMCC Frequency
Resolution

• Objective Improve the frequency resolution of
  the signal detected
• Signals like microbaroms are detected by
  DFX-PMCC at frequencies up to 1Hz
• Action Use a band-pass filter that has a steep
  roll off
• Results
• Slightly decreases the detection capability but
  substantially improves the frequency resolution
• The microbaroms detected by DFX-PMCC are more
  constrained to lower frequencies lt0.5Hz

12

Baseline Configuration for DFX-PMCC Frequency
Resolution
Examples of microbaroms detected at I07AU (6
Sep2003)
DFX-PMCC, Chebychev order 2
I07AU
DFX-PMCC, Chebychev order 3 (Steeper roll-off)
13

Baseline Configuration for DFX-PMCC Azimuth
Resolution

• Objective Improve the azimuth resolution
• Trigger detections on signals with stable
  characteristics, focus on azimuth stability
• Prepare for a basic event categorization
• Action
• Use a low threshold in azimuth when forming the
  PMCC families (/- 6 or 10 degrees)
• Find a good compromise between a good azimuth
  resolution and high number of detections produced
  (split families). Risk of merging independent
  families if the azimuth criterion is too loose.

Unstable detections
Azimuth threshold
Event categorization
Split families
(Total of detections)
Azimuth accuracy
Missed small detections
14

Baseline Configuration for DFX-PMCC Azimuth
Resolution
Results Example of microbaroms at I07AU (6 Sep
2003)
Parameter Daz /- 10 deg Output 55
detections backaz 225.5 /- 7.7
Parameter Daz /- 6 deg Output 89
detections backaz 224.7 /- 5.4
15
Baseline Configuration for DFX-PMCC Array Geometry

• Objective
• Study how the selection of array geometry
  impacts on the detection capability
• Results presented at the IMS infrasound meeting
  in Vienna, March 2003
• Action
• For identified signals (Concorde, microbaroms)
  check the detections when removing one array
  element from the processing
• Results
• Good detection capability and quality of PMCC
  even with low number of array elements (4)
• Good detection capability of small aperture
  arrays well complemented by quality of detection
  features provided by large aperture arrays
• The central element plays a key role in the PMCC
  detection processing
• DFX-PMCC shall configuration includes a large
  number of combinations of sub-networks
• Plan intelligent selection of sub-networks
  based on data channel QC (Future evolution)

16
Baseline Configuration for DFX-PMCC Array Geometry
Example I34MN 18 Jan 2003 microbaroms between
800-1200 GMT
17

Baseline Configuration for DFX-PMCC Detection
Threshold

• Objective Set the minimum size of events to be
  detected
• DFX-PMCC does not work with amplitudes (no
  minimum SNR threshold applied)
• The size of the signal to be detected depends
  on the minimum number of pixels per family
• It is the last parameter to be tuned according
  to the types of events detected at the station

Family would not be built if min. number of
pixelslt5
18

Baseline Configuration for DFX-PMCC Detection
Threshold

• Example Argyle mine at I07AU, ML2.4 730km
  (16Apr2002)

Pixel threshold15 -gt 13 detections
Pixel threshold100 -gt 1 detection
Pixels
Pixels
Families
Families
SNR9.8
19

DFX-PMCC testing and validation
Test the baseline configuration of DFX-PMCC on
the 30 IDC infra. ref. events
2001 SEL3 DFX-PMCC Day Event
description IMS array Seismic Infra Seismic Infra
23 Apr North Pacific bolide I10CA x
x I26DE x x I59US x a 23 Jun Peru
main earthquake I08BO x x a a 23 Jun Peru
aftershock I08BO x x r a 7 Jul Peru
earthquake I08BO x x x x 15 July Etna
eruption I26DE r x 21 Sep Toulouse
explosion I26DE a a 14 Oct Canada bolide I10CA
(!) r x 21 Oct Nuremberg explosion I26DE
r a 14 Nov China earthquake I34MN a a a a 18
Nov Leonid meteorite I59US x a 2002 26
Jan US missile test I59US r a 9 Mar Central
Pacific bolide I10CA (!) a a I57US x
a I59US a a 12 Mar Space Shuttle
STS109 I10CA (!) r a 6 Apr Germany
bolide I26DE a a
20

DFX-PMCC testing and validation
Test the baseline configuration of DFX-PMCC on
the 30 IDC infra. ref. events
2002 (Cont.) SEL3 DFX-PMCC Day Event
description IMS array Seismic Infra Seismic Infra
16 Apr Argyle mine I07AU a a 6
Jun Mediterranean sea bolide I26DE a a 8
Sep New Guinea earthquake I07AU a r 28
Sep Watusi explosion I10CA (cor.) r a 28
Sep Indonesia earthquake I07AU a a a a 23
Oct Alaska earthquake I10CA (cor.) x a x
a 3 Nov Alaska earthquake I10CA (cor.) a a
a a 2003 1 Feb Space Shuttle STS107 I10CA
(cor.) a a 1 Feb Concorde I26DE r a 27
Mar Explosion Billy-Berclau I26DE r a 8
Jun Soyuz rocket launch I31KZ a a 12
Aug Soyuz rocket launch I31KZ a a 14
Aug Greece earthquake I26DE r a 27
Aug Hawaii earthquake I59US a a 29 Aug Soyuz
rocket launch I31KZ a a 15 Oct LongMarch
rocket launch I34MN a a
21

Summary

• History of DFX-PMCC connected to
• The amount of infrasound data processed at the
  IDC ( of IMS stations)
• Large contribution of WinPMCC (new vision for
  presenting the derived results)
• DFX-PMCC provides a powerful automatic detector
  for infrasound signals with
• Good detection capability and feature extraction
  quality
• Large variety of signals detected
• Low level of false alarms
• DFX-PMCC brings reliability into IDC infrasound
  processing
• Since September 2003, Maturity of the software
  and high level of confidence
• The baseline configuration has been successfully
  tested on the IDC reference events
• Plans for the near future
• Large scale testing in a real-time environment
  is to be initiated (November 2003)
• Limit the impact of long duration signals on the
  automatic processing
• Develop in parallel the future interactive
  infrasound software at the IDC