1 / 17

Infrasound Data Processing at the PIDC

Infrasound Data Processing at the PIDC. Greg Beall Robert Woodward Science Applications International Corporation Center for Monitoring Research. Outline. Background Detection and feature estimation Phase identification Event formation Interactive review Current status Challenges

hume
Télécharger la présentation

Infrasound Data Processing at the PIDC

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Infrasound Data Processing at the PIDC Greg Beall Robert Woodward Science Applications International Corporation Center for Monitoring Research

  2. Outline • Background • Detection and feature estimation • Phase identification • Event formation • Interactive review • Current status • Challenges • Resources at CMR 2

  3. Background • Work on infrasound data processing for the PIDC started in Spring 1996 • Automatic infrasound detection and event formation at PIDC began in Spring 1998; interactive analyst review began in late 1998 • Delivered to IDC in Release 1 (Spring 1998) • Most development work was tested with a small regional network of short and intermediate baseline stations provided by the DOE • Release 3 delivered to the IDC in December, 2000 • Release 3 Upgrade delivered to the IDC in September, 2001 • The PIDC experiment concluded on September 30, 2001 • The Center for Monitoring Research will continue to support the infrasound research and development community through • collection, archive, and distribution of infrasound data • infrasound waveform library • infrasound resource guide 3

  4. Infrasound Signal Detection • Traditional approach is based on spatial coherence alone • Test statistic, e.g. normalized cross-correlation (NCC), or lag-closure discrepancy • Subject to capture by continuous clutter (microbaroms, mountain-associated waves) • PIDC Infrasound Detector (DFX/libinfra) • Detects based on clusters in slowness plane • Allows detection of a contemporary off-azimuth signals in the presence of clutter • Coincidence detector requires time-coincident spatial coherence and excess energy • f-statistic plus STA/LTA to declare coincident detection • Suppresses detections due to continuous clutter as well as wind bursts • Configured via beam recipes and detector parameters • Beam recipes specify azimuth/slowness steering, frequency bands, sensor groups, and trigger thresholds • Parameters specify coherent detector duration and overlap, STA/gap/LTA durations, detection thresholds (coherence and energy) 4

  5. Estimated Features • Detection table • Time, azimuth, slowness, and uncertainties; fstat • Used in association and location • Amplitude table • Amplitude, period • Has potential value for characterizing event, determining propagation path, restricting incorrect associations • Infra_features table • Energy, coherence, coincidence start time and duration • Zero-crossing and corner frequencies • Coherence period, coherence SNR, and total energy • Measurements available to characterize signals, but utility has not been demonstrated yet 5

  6. Phase Identification / Station Processing • Phase Identification is important for event formation and location • The Global Association process (GA) uses phase ID to restrict association • Locator uses phase ID to select travel-time curve • Also used in event confirmation tests • Phase Identification is done by StaPro (station processing) • Currently categorizes phase as I, Ix, or N (i.e. first infra, later infra, noise) • Distinction between I and N based on slowness • I-type are grouped by time and azimuth • First phase in a group labeled I, remainder labeled Ix • Would like to distinguish phases further (e.g. stratospheric, thermospheric) to better restrict travel times • Seismic identifies Pn, Pg, Sn, Lg, Rg, P, S; N, coda • Uses signal and contextual information • Routinely compute infrasound features to support characterization • Need established signals which can be characterized 6

  7. Network Processing - GA • GA processes detections and features computed by DFX and StaPro applications for a network of stations to form events • DFX performs detection and feature extraction • Exhaustive grid search is conducted to hypothesize events and associate detections • Associations must satisfy constraints on time, azimuth and slowness residuals • Conflict resolution removes non-unique associations • Final event must satisfy consistency and confirmation tests • GA runs in 3 pipelines; delayed 2, 4 and 10 hours • GA processes overlapping windows of data every 20 minutes • Infrasound arrivals are processed with the other waveform technologies • Infrasound arrivals are only associated in 2nd and 3rd pipelines 7

  8. Automatic Infrasound Phase Association • Associations are based on phase ID, time, azimuth, and slowness • Only I phases are used to compute location • Only one I phase can be associated per station • Time, azimuth and slowness residuals must be consistent with combination of measurement and modeling errors • Minimum event criteria require infrasound arrivals at two stations • Mixed-technology events are allowed, but are restricted to limit false alarms • Infrasound arrivals can be associated with events detected by other technologies but can not be used to corroborate an event detected seismically 8

  9. Travel-Time Tables • Travel-time tables are used to restrict phase associatiation and to locate events • Same tables used in automatic event formation and interactive review • Only a single I phase is specified • Although could apply distinct travel-time tables for different paths (e.g. stratospheric and thermospheric), can't currently distinguish phases using signal characteristics alone • Utilize either global 1-D table or station-specific 2-D tables • 2-D tables include average seasonal wind variation • Computed from MSISE-90 atmospheric model and HWM-93 seasonal wind model (Hedin at al., 1991, 1996) • 1-D and 2-D tables use different modeling errors 9

  10. Analyst Review of Infrasound Data • PIDC analysts interactively reviewed all automatically associated infrasound phases • PIDC analysts did not generally scan for missed infrasound detections or missed infrasound associations • PIDC analysts used ARS to review amplitude and coherence traces and used XfkDisplay to confirm azimuth • Also had access to azimuth and slowness traces for confirmation • Event confirmation more stringent; required at least 3 azimuths and 2 times • Vast majority of automatic associations were removed by analysts 10

  11. Current Status/Observed Behavior • Relatively unchanged since R3 • Main changes were to eliminate duplicate detections caused by window overlap and to provide 2-D seasonal travel-time tables for all IMS stations • Average statistics for July - September, 2001 • 12 stations: IS08, IS10, IS26, IS34, IS55, IS57, IS59; DLIAR, NVIAR, PDIAR, SGAR, WRAI • About 20 detections/day/station at 12 stations, 60% I • Some station averages much higher, e.g. IS55 and IS57 about 70 detections/day • About 2.6 automatic associations/day; none retained in REB • Expert review of automatic detection indicates that • Generally detect most significant signals • Accurate measure of azimuth and slowness and their measurement errors for these detections • Occasional missed detections • Numerous microbarom detections • Summary since 1998 • Total of six REB events with associated infrasound signals • Three REB events were purely infrasound (4/23/01) 11

  12. Challenges Going Forward • Current Limitations • Most of our experience has been with short-baseline arrays in North America • Detector is reasonably stable and robust, but has not had the scrutiny of seismic system • Event formation and location is rudimentary and has had very little realistic testing due to the sparse network that has been available 12

  13. Challenges Going Forward (2) • Signal Detection at New Stations • New longer baseline arrays will need to be tuned and may suffer from spatial aliasing in higher frequency bands • Detector may not be optimal for processing multiple baseline arrays • assumes single coherence integration and STA/gap/LTA durations for all beams (filter bands, sensor groups) • Detector assumes planar configuration • Can cause problems for stations like IS59 • Noise environment at new stations, especially oceanic, may require changes in strategy for handling noise clutter 13

  14. Challenges Going Forward (3) • Automatic Event Processing • A denser network will increase the frequency of false alarms along with true event detection • Can simulate the impact under different assumptions • SynGen, delivered in the R3 Upgrade, can be used for this • Problem can be mitigated by using better modeling and signal characterization to restrict association • Use of distinct phase IDs for tighter travel time checks • Amplitude or spectral consistency checks • Correction of azimuth, amplitude for wind • Well located reference events will be essential for evaluating the validity and utility of refined signal and propagation models • Interactive Processing • Will need to adapt practices to accommodate new data 14

  15. Resources for Infrasound Research at CMR • Infrasound data • IMS arrays • DOE arrays • See new Infrasound Resource Guide for details • Array descriptions • Data availability • Data access • Event information Array Descriptions Data Availability 15

  16. ALL Auroral Wave Chemical Explosion Earthquake Explosion Gas Pipe Explosion Gravity Wave Microbaroms Mountain Waves Nuclear Explosion Rocket Launch Synthetic Unknown Volcano ALL GROUND TRUTH MAXWELL UofA/ENSCO SYNTHETIC Resources for Infrasound Research (2) • Infrasound waveform library (www.cmr.gov/rdss/) • University of Alaska/ENSCO -- miscellaneous signals • Maxwell (SAIC) -- collection of recordings of Soviet atmospheric nuclear tests • Ground truth events • LANL synthetics • Others 16

  17. Summary • The Release 3 Upgrade system was delivered to the IDC in September, 2001 • The R3U system utilizes a highly tunable infrasound processing system that detects infrasound signals and associates them with other S/H/I signals to form and locate events • New IMS infrasound data will allow (and probably necessitate) refinement of the current processing • Detector tuning • New array geometries and baselines • New site-specific noise backgrounds • Denser network • Will allow more events to be formed; more false alarms as well • Phase identification and travel-time/attenuation models • Refined analysis procedures • CMR will continue to support US infrasound R&D community • IMS and non-IMS data • IDC products • Waveform collections • Assistance in accessing and utilizing CMR resources 17

More Related