Academia.eduAcademia.edu

Figure 4 – uploaded by Ronan Le Bras

See full PDFdownloadDownload figure
Table 8: Estimated Event Information for the two WSMR shots. Interesting observations include that wavefronts with slower phase velocity arrived earlier than wavefronts with higher phase velocities at LSAR, indicating that the later rays reached higher altitudes and sampled higher velocities.

Table 8 Estimated Event Information for the two WSMR shots. Interesting observations include that wavefronts with slower phase velocity arrived earlier than wavefronts with higher phase velocities at LSAR, indicating that the later rays reached higher altitudes and sampled higher velocities.

Related Figures (44)

This is illustrated by Table 2 which summarizes the results of tests on 15-day synthetic data sets with three different levels of noise. The first is noise-free, the second has noise added at an average rate of 2 arrivals per day (called 12 hour random) and the third has noise added at a rate of 8 arrivals per day (called 3 hour random). There were 10 target events in the synthetic data set with 3 or more defining phases. Table 2: Simulations on synthetic hydroacoustic detections with varying numbers of random detections and precision on azimuth measurement.
This is illustrated by Table 2 which summarizes the results of tests on 15-day synthetic data sets with three different levels of noise. The first is noise-free, the second has noise added at an average rate of 2 arrivals per day (called 12 hour random) and the third has noise added at a rate of 8 arrivals per day (called 3 hour random). There were 10 target events in the synthetic data set with 3 or more defining phases. Table 2: Simulations on synthetic hydroacoustic detections with varying numbers of random detections and precision on azimuth measurement.
Figure 2. Map of travel time differences between the summer travel times for station WK31 and a constant velocity model (1485 m/s).
Figure 2. Map of travel time differences between the summer travel times for station WK31 and a constant velocity model (1485 m/s).
The travel time handling system has the capability to use station-specific and path-dependent travel times calculated using average seasonal wind patterns and we are in the process of integrating such tables for operational stations into the PIDC_7.0 release. These travel time tables were provided to us by D. Brown from the Center for Monitoring Research. An example of travel time map is shown in Figure 3.
The travel time handling system has the capability to use station-specific and path-dependent travel times calculated using average seasonal wind patterns and we are in the process of integrating such tables for operational stations into the PIDC_7.0 release. These travel time tables were provided to us by D. Brown from the Center for Monitoring Research. An example of travel time map is shown in Figure 3.
Figure 7. (from Brumbaugh and Le Bras, 1998). Location of the 141 individual explosions superimposed on bathymetric data and the coastline of northern Honshu, Japan. The smaller circles indicate explosions with a yield of 20 kg, the intermediate circles represent yields of 40 kg and the two large circles indicate 400 kg explosions. September 1996 Refraction Survey Explosion Locations
Figure 7. (from Brumbaugh and Le Bras, 1998). Location of the 141 individual explosions superimposed on bathymetric data and the coastline of northern Honshu, Japan. The smaller circles indicate explosions with a yield of 20 kg, the intermediate circles represent yields of 40 kg and the two large circles indicate 400 kg explosions. September 1996 Refraction Survey Explosion Locations
Figure 9. Histograms of travel time residuals (observed minus predicted travel times) computed at WAKE and PSUR using both the 1D constant velocity model and the 2D travel time tables. Note the vertical scale on the histograms differs between WAKE and PSUR. A significant improvement in travel time accuracy is obtained when the 2D travel time tables are employed, particularly for PSUR detections.
Figure 9. Histograms of travel time residuals (observed minus predicted travel times) computed at WAKE and PSUR using both the 1D constant velocity model and the 2D travel time tables. Note the vertical scale on the histograms differs between WAKE and PSUR. A significant improvement in travel time accuracy is obtained when the 2D travel time tables are employed, particularly for PSUR detections.
Cepstral Analysis for Sept 1998 explosions Figure 11. Bubble pulse delays computed from the cepstrum at each station for the September 1998 refraction experiment. The delay times vary from explosion to explosion but are extremely consistent between the three stations. The step-like appearance is due to the limited resolution (2 samples or 0.006 seconds) on the delay estimation.
Cepstral Analysis for Sept 1998 explosions Figure 11. Bubble pulse delays computed from the cepstrum at each station for the September 1998 refraction experiment. The delay times vary from explosion to explosion but are extremely consistent between the three stations. The step-like appearance is due to the limited resolution (2 samples or 0.006 seconds) on the delay estimation.
Figure 12. These rose diagrams show for each 10 degree bins a quantity computed from the number of REB events with an association only at one of WK30 (wk30) or WK31 (wk31) or at both stations (both). Bins with fewer than 15 events are not included in the display for lack of sufficient statistics. For station WK30, the However, WK31 does not display an area of complete blockage, and the expected shadow of the island and related bathymetric heights is not seen. This may be due to the closer proximity of the WK30 hydrophone to Wake island.
Figure 12. These rose diagrams show for each 10 degree bins a quantity computed from the number of REB events with an association only at one of WK30 (wk30) or WK31 (wk31) or at both stations (both). Bins with fewer than 15 events are not included in the display for lack of sufficient statistics. For station WK30, the However, WK31 does not display an area of complete blockage, and the expected shadow of the island and related bathymetric heights is not seen. This may be due to the closer proximity of the WK30 hydrophone to Wake island.
Table 4: Ground-Truth of two WSMR Explosions. The Los Alamos, New Mexico (LSAR) and Lajitas, Texas (TXIAR) infrasound arrays are in the proximity of WSMR as shown in Figure 14. The expected distance and azimuth for the two stations relative to the two WSMR shots are shown in Table 5. Infrasound energy wavetrains directly attributable to the shots span a time interval of 4 to 5 minutes after the first arrival. The travel-time of the first arrival to LSAR and TXIAR was about 14 minutes and 27 minutes, respectively.
Table 4: Ground-Truth of two WSMR Explosions. The Los Alamos, New Mexico (LSAR) and Lajitas, Texas (TXIAR) infrasound arrays are in the proximity of WSMR as shown in Figure 14. The expected distance and azimuth for the two stations relative to the two WSMR shots are shown in Table 5. Infrasound energy wavetrains directly attributable to the shots span a time interval of 4 to 5 minutes after the first arrival. The travel-time of the first arrival to LSAR and TXIAR was about 14 minutes and 27 minutes, respectively.
Table 5: Ground-Truth Station Information. At the LSAR array we identified four distinct arrivals that can be correlated between the two WSMR shots with similar timing and phase velocity within the sequence. The larger shot (November 19) has additional arrivals beyond these four. Four arrivals are also discernible at the TXIAR array for both shots, however the correlation of the timing between the shots is not as good as at LSAR. The quality of the arrivals at TXIAR is not as high as at LSAR due to the longer distance to TXJAR.
Table 5: Ground-Truth Station Information. At the LSAR array we identified four distinct arrivals that can be correlated between the two WSMR shots with similar timing and phase velocity within the sequence. The larger shot (November 19) has additional arrivals beyond these four. Four arrivals are also discernible at the TXIAR array for both shots, however the correlation of the timing between the shots is not as good as at LSAR. The quality of the arrivals at TXIAR is not as high as at LSAR due to the longer distance to TXJAR.
Approximate locations for both shots were estimated using first arrivals from the LSAR and TXIAR arrays. A constant velocity model of 320 m/s was used to derive these locations. Information for the two associated arrivals for the November 12 shot are listed in Table 6, while Table 7 lists the attributes for the two associated arrivals for the November 19 shot.
Approximate locations for both shots were estimated using first arrivals from the LSAR and TXIAR arrays. A constant velocity model of 320 m/s was used to derive these locations. Information for the two associated arrivals for the November 12 shot are listed in Table 6, while Table 7 lists the attributes for the two associated arrivals for the November 19 shot.
Figure 15. Estimated locations of November 12 and November 19 1997 explosions. First arrivals from the LSAR and TXIAR infrasonic arrays were used to form this location. Azimuth and time were defining at both stations for both shots. The 90% error ellipses are shown centered about the estimated locations. The actual locations, very close to each other, are represented by the star. The event estimated location for the November 12 shot (first shot) was found to be 15 km south- southwest from the actual location and the origin time is 26 seconds late. For the November 19 shot (second shot), the estimated location was found to be about 27 km south-southeast from the actual location and the origin time was estimated to be 18 seconds early. The estimated events information is listed in Table 8 and the locations are shown on Figure 15. Note that the size of the error ellipses on that figure reflect a larger modeling error than is currently used in operations. A more detailed description of the data associated with these two shots can be found in Jenkins et al.(1997).
Figure 15. Estimated locations of November 12 and November 19 1997 explosions. First arrivals from the LSAR and TXIAR infrasonic arrays were used to form this location. Azimuth and time were defining at both stations for both shots. The 90% error ellipses are shown centered about the estimated locations. The actual locations, very close to each other, are represented by the star. The event estimated location for the November 12 shot (first shot) was found to be 15 km south- southwest from the actual location and the origin time is 26 seconds late. For the November 19 shot (second shot), the estimated location was found to be about 27 km south-southeast from the actual location and the origin time was estimated to be 18 seconds early. The estimated events information is listed in Table 8 and the locations are shown on Figure 15. Note that the size of the error ellipses on that figure reflect a larger modeling error than is currently used in operations. A more detailed description of the data associated with these two shots can be found in Jenkins et al.(1997).
Table 9: Arrival Information For Micare Mine Explosion. The location obtained using the seismic arrivals alone is slightly to the north of the final location estimated using all three Pn, Lg, and J arrivals. The addition of the J arrival improves the azimuth, reduces the error, and moves the location farther to the south where it overlaps the mining area (Figure 21). The high phase velocity is compatible with the interpretation of the lone infrasound phase for this event as a thermospheric arrival. This is further confirmed by its travel time of 20:18.3 (or 1218.3 sec) at a distance of 303 km. The location is determined using a constant velocity of 320 m/s for the infrasound waves in the atmosphere, but this results in a positive time residual of 243.24 s. One large infrasound arrival is recorded and two main seismic arrivals, Pn and Lg, are clearly observable and used in concert to form a location (Figures 19 and 20). The infrasound arrival shows high coherency across all four channels of the array. The measured travel-time for the I phase to reach TXIAR is approximately 20 minutes and 18 seconds. The picked arrival information used in the joint seismic-infrasound location is listed in Table 9 for both the seismic and infrasound phases.
Table 9: Arrival Information For Micare Mine Explosion. The location obtained using the seismic arrivals alone is slightly to the north of the final location estimated using all three Pn, Lg, and J arrivals. The addition of the J arrival improves the azimuth, reduces the error, and moves the location farther to the south where it overlaps the mining area (Figure 21). The high phase velocity is compatible with the interpretation of the lone infrasound phase for this event as a thermospheric arrival. This is further confirmed by its travel time of 20:18.3 (or 1218.3 sec) at a distance of 303 km. The location is determined using a constant velocity of 320 m/s for the infrasound waves in the atmosphere, but this results in a positive time residual of 243.24 s. One large infrasound arrival is recorded and two main seismic arrivals, Pn and Lg, are clearly observable and used in concert to form a location (Figures 19 and 20). The infrasound arrival shows high coherency across all four channels of the array. The measured travel-time for the I phase to reach TXIAR is approximately 20 minutes and 18 seconds. The picked arrival information used in the joint seismic-infrasound location is listed in Table 9 for both the seismic and infrasound phases.
Table 10: Arrival Information For Tyrone Explosion. The emergent or packet-like appearance of the infrasound arrivals made them difficult to window for the F-K analysis and while we observed 7 individual detections (shown in Figure 23 and listed in Table 10) within the wavetrain only 4 of them produced acceptable, if rather low quality, F-K results (those with snr=fair). The optimal bandpass filter for the infrasound channels is found to be 0.8-3.0 Hz, and once again we experimented with the lead/lag window time and bandwidth parameters of the F-K analysis. The ground-truth arrivals found for this station-event pair are listed in Table 10 Both seismic phases Pn and Lg are clearly observed and contain high frequency energy as seen in Figure 23 where the trace is filtered between 3.0 and 6.0 Hz.
Table 10: Arrival Information For Tyrone Explosion. The emergent or packet-like appearance of the infrasound arrivals made them difficult to window for the F-K analysis and while we observed 7 individual detections (shown in Figure 23 and listed in Table 10) within the wavetrain only 4 of them produced acceptable, if rather low quality, F-K results (those with snr=fair). The optimal bandpass filter for the infrasound channels is found to be 0.8-3.0 Hz, and once again we experimented with the lead/lag window time and bandwidth parameters of the F-K analysis. The ground-truth arrivals found for this station-event pair are listed in Table 10 Both seismic phases Pn and Lg are clearly observed and contain high frequency energy as seen in Figure 23 where the trace is filtered between 3.0 and 6.0 Hz.
Table 13: Average Residuals Between ‘“‘Ground-Truth” and GA Events We computed the ‘esiduals between simulations (Table with random detec 10 simulations wi average, standard deviation, and median distance, depth, and origin time he ground-truth explosions and the GA events linked to them for the two sets of 13). The average distance and origin time residuals for the five simulations ions are an order of magnitude larger than the same residuals computed for the thout random detections. This observation is not surprising because of the additional number of random arrivals present in the five simulations which could be misassociated ‘o events and thereby skew the final hypocenter and origin time. The depth residuals do not vary nuch between simulation sets - most of the GA events linked to ground-truth explosions were dlaced at the surface in both simulation sets (the ground-truth explosions were placed at depths of ).2 km).
Table 13: Average Residuals Between ‘“‘Ground-Truth” and GA Events We computed the ‘esiduals between simulations (Table with random detec 10 simulations wi average, standard deviation, and median distance, depth, and origin time he ground-truth explosions and the GA events linked to them for the two sets of 13). The average distance and origin time residuals for the five simulations ions are an order of magnitude larger than the same residuals computed for the thout random detections. This observation is not surprising because of the additional number of random arrivals present in the five simulations which could be misassociated ‘o events and thereby skew the final hypocenter and origin time. The depth residuals do not vary nuch between simulation sets - most of the GA events linked to ground-truth explosions were dlaced at the surface in both simulation sets (the ground-truth explosions were placed at depths of ).2 km).
Table 14: Average Number of Detections
Table 14: Average Number of Detections
Figure 30. (from Hanson et al., 2000). REB events with T phases between May 1997 and May 1999. There are 8,437 events in the two year period that have an observed T phase at a least one hydrophone station. Most of the events are along the western edge of the Pacific. There are a few events that appear questionable, but they compose a very small percentage of the total (< 0.1%).
Figure 30. (from Hanson et al., 2000). REB events with T phases between May 1997 and May 1999. There are 8,437 events in the two year period that have an observed T phase at a least one hydrophone station. Most of the events are along the western edge of the Pacific. There are a few events that appear questionable, but they compose a very small percentage of the total (< 0.1%).
Table 15: Performance of Automatic Processing
Table 15: Performance of Automatic Processing
Table 16: Average Number of Detections Table 16 shows the detection and association statistics for the whole year. The propagation model used in the association process is a constant velocity model (320 m/s). An automatic event may be formed with a minimum of two infrasound arrivals. We are presenting the results of the automatic processing in Figure 31 for events with two or more associated J arrivals and Figure 32 for events with three or more associated arrivals. Most of these events have not been reviewed by analysts o1 infrasound specialists and we have no definite criteria to label these events as real or false alarms, however a review of a similar set of events was performed at the CMR (Gault and Brown, 1999) where it was concluded that 20 to 30% of the events they looked at were real transient events, although no ground truth is available to confirm their location and time of origin. It is likely that some of these events originate from mining or quary blasts or ordinance disposal.
Table 16: Average Number of Detections Table 16 shows the detection and association statistics for the whole year. The propagation model used in the association process is a constant velocity model (320 m/s). An automatic event may be formed with a minimum of two infrasound arrivals. We are presenting the results of the automatic processing in Figure 31 for events with two or more associated J arrivals and Figure 32 for events with three or more associated arrivals. Most of these events have not been reviewed by analysts o1 infrasound specialists and we have no definite criteria to label these events as real or false alarms, however a review of a similar set of events was performed at the CMR (Gault and Brown, 1999) where it was concluded that 20 to 30% of the events they looked at were real transient events, although no ground truth is available to confirm their location and time of origin. It is likely that some of these events originate from mining or quary blasts or ordinance disposal.
Conceptual View of Fusion Architectures (from Hall 1992)
Conceptual View of Fusion Architectures (from Hall 1992)
Related topics:
Signal ProcessingUnderwater AcousticsRadioactive Isotopes
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved