Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Fig 1.

Proposed framework for P-wave detection and picking that accommodates any modeling technique (for prediction) and a t-f tool (for decomposition).

More »

Fig 1 Expand

Table 1.

List of parameters used by the algorithm with brief description and values used in the study.

More »

Table 1 Expand

Fig 2.

Schematic of MODWPT decomposition upto level 4.

‘A’ and ‘D’ refers to approximation and detail respectively. Integer in the top-right corner of each box indicates the packet number in increasing frequency order.

More »

Fig 2 Expand

Table 2.

Frequency mapping for MODWPT packets.

Level indicates the level of decomposition and the frequencies are reported for a signal sampled at 20 Hz.

More »

Table 2 Expand

Table 3.

Details of parameters used to generate synthetic data.

More »

Table 3 Expand

Fig 3.

Low SNR synthetic data with a short-lived event embedded in noise.

More »

Fig 3 Expand

Fig 4.

Packet selection.

The plot on top depicts the selected packets on the fly based on the ranking of highlighted packets. Y-axis represents the window onset time in the ranked order (top to bottom). The presence of 1 indicates detection, and 0 represents no detection. The duration of highlighted packets is shown in the bottom plot, where the X-axis represents the onset of the working window.

More »

Fig 4 Expand

Fig 5.

μAD in the highlighted packets.

There is a sudden increase in the μAD in certain packets (30, 14 and 17) while in other packets (16 is shown in the figure, but it holds for the rest of the packets), μAD is always below the threshold.

More »

Fig 5 Expand

Fig 6.

Reconstructed signal.

Amplitude of the reconstructed signal is smaller than the original s[k].

More »

Fig 6 Expand

Fig 7.

Selected packets for different SNR cases.

Number of highlighted packets increases as the SNR increases.

More »

Fig 7 Expand

Table 4.

Seismic data.

Details of different datasets downloaded from IRIS.

More »

Table 4 Expand

Fig 8.

Event detection results for low SNR datasets.

Plots (a), (d), (g) shows the vertical channel data from MAJO, ANTO, and ANMO station respectively, where the vertical black solid line indicates the onset of true event as reported on the IRIS website, plots (b), (e), (h) depicts corresponding μAD in the selected packets. The solid horizontal line in the middle plots is the threshold value for the respective packets. Zoomed snapshots of respective data with picked events is shown in plots (c), (f) and (i) where the red vertical lines indicate the time at which the proposed method picks the event. Plots (c), (f), (i) shows the zoomed snapshot of data with picked events.

More »

Fig 8 Expand

Fig 9.

ANMO station: Event detection vs SNR.

Top and bottom figures depict the detection error and SNR, respectively.

More »

Fig 9 Expand

Fig 10.

Detection rate for low SNR conditions.

Detection rate of the proposed algorithm for SNR <5 dB. These datasets have events in the magnitude range from 0.13 to 4.

More »

Fig 10 Expand

Fig 11.

Detection error.

The left plot depicts the detection rate of the proposed algorithm for high SNR >1 dB. The right plot shows the false alarm rate of the proposed method for background noise, where the red square box indicates no detection in the noise and blue boxes indicate false detection of events in background noise.

More »

Fig 11 Expand

Fig 12.

Low magnitude event detection results.

Vertical channel TSUM station (a) data with two events—solid vertical black line indicates the true onset of the seismic event of interest and the dashed black color line indicates the second event in the data. Plots (b), (c) and (d) depict the μAD corresponding to packets 9, 16 and 27 respectively and plot (e) is a zoomed snapshot of data—the red vertical line indicate the time at which the proposed algorithm picks the low magnitude event.

More »

Fig 12 Expand

Fig 13.

Comparison with existing detector.

Detection error in samples of STA/LTA and the proposed method for events of different magnitude range.

More »

Fig 13 Expand

Fig 14.

Comparison with existing pickers.

Picking error in samples is reported for all the four categories, (a) Mag ≤2.5, (b) 2.5< Mag ≤4, (c) 4< Mag ≤6 and (d) Mag >6.

More »

Fig 14 Expand

Fig 15.

False alarm rate for different P-wave arrival detector / picker.

More »

Fig 15 Expand

Table 5.

Detection/Picking rate and false alarm rate in percentage.

FAR stands for false alarm rate which is defined as the % of datasets with false detection of event.

More »

Table 5 Expand

Fig 16.

Sensitivity of crucial parameters.

Effects of (a) varying window length for a fixed sliding length of 5 samples and (b) varying sliding length for fixed window length of 240 samples on the performance of the proposed method. Vertical axis indicates the detection error for varying tuning parameters. Blue color circles indicate the detection error, and the red color solid line is the fitted curve for both the parameters.

More »

Fig 16 Expand