Labelled set of waveforms from the Borovoye Geophysical Observatory nuclear explosion archive for signal reconstruction testing

Steven J. GibbonsAugust 14, 2025 An et al. (2015) describe in detail an archive of nuclear explosion signals recordedbetween 1966 and 1996 at the Borovoye Geophysical Observatory in Kazakhstan.The original waveforms are made openly available onhttps://www.ldeo.columbia.edu/res/pi/Monitoring/Data/ The waveforms are recorded with different gains such that some of them are"complete" (capturing the full dynamic range of the signals) while others are clipped.I have taken many of the waveforms from this archive and, as best I can, labelled whichsamples are clipped. It is my intention that this collection of annoated waveforms can be usedas a testbed for the reconstruction of clipped waveforms.There are many papers in the literature that address this subject and I plan toexplore the applicability and effectiveness of different methods. I am open to thepossibility (probability) that a meaningful reconstruction is impossible for manyof the signal segments here, and am equally convinced that many signals can almostcertainly be restored using existing methods. The folder ASCII_ARCHIVE contains the selected waveforms written out in an ASCII formatfor easy visualization, and flagged to the best of my ability (using a few simple pythonalgorithms) which samples are "correct" and which are clipped. The format of the filesis as follows (taking the file 19670922.0504.brvk.KODB.SLZb0.030.txt as an example): # 1967-09-22T05:04:00.000 0.03000 10997 -71866560.00000 BRVK_SLZb 0 70.437988 0 1 -0.405301 0 2 -1.486040 0 3 -3.619660 0 4 -6.067526 0 5 -8.281006 0 6 -10.185470 0 7 -11.560674 0 8 -9.553430 0 9 -9.328013 0 10 -8.562012 0 11 -6.562012 0... The first line givesthe time of the first sample of the time-series, T0,the sampling interval in seconds, DT,the number of samples present, NSAMP,the epoch time (relative to 1970-00-00Z00:00:00.000) of the first sample,and the name of the channel as recorded in the original archive.Each of the NSAMP subsequent lines contains 3 numbers, ISAMP, RVAL, and ICLIP. ISAMP is the number of the sample, such that the time of the sample is T0 + DTISAMP*.RVAL is the value of the original data at this sample.ICLIP is set to: zero if the true value is likely well represented by RVAL, -1 if the true value is likely less than RVAL (often the negative clipping value), and +1 if the true value is likely greater than RVAL (often the positive clipping value). Many of the flags may be eroneous and spurious values between the clippinglimits are likely to be unflagged. The complete list of files is given in the file BRVK_ASCII_file_summary.txt. The first few lines of this file read: 1966-12-18T04:57:30.000 0.03000 44992 -95886150.00000 7 88 6 BRVK_SHE 19661218.0457.brvk.KODB.SHE0.030.txt1966-12-18T04:57:30.000 0.03000 44977 -95886150.00000 8 86 7 BRVK_SHN 19661218.0457.brvk.KODB.SHN0.030.txt1966-12-18T04:57:30.000 0.03000 44952 -95886150.00000 9 83 9 BRVK_SHZ 19661218.0457.brvk.KODB.SHZ0.030.txt1966-12-18T04:57:30.000 0.03000 44975 -95886150.00000 2 98 2 BRVK_SLZb 19661218.0457.brvk.KODB.SLZb0.030.txt1967-02-26T03:58:30.000 0.03000 34996 -89841690.00000 1 98 2 BRVK_SLZb 670226.0358.brvk.KODB.SLZb0.030.txt The first four terms are as described above.We then have 3 rounded up percentages: %negativeclipped, %notclipped %positiveclipped(so 0 100 0 means a perfect waveform)followed by the channel and the filename. References An, V. A., Ovtchinnikov, V. M., Kaazik, P. B., Adushkin, V. V., Sokolova, I. N., Aleschenko, I. B., Mikhailova, N. N., Kim, W.-Y., Richards, P. G., Patton, H. J., Scott Phillips, W., Randall, G., & Baker, D. (2015).A digital seismogram archive of nuclear explosion signals,recorded at the Borovoye Geophysical Observatory, Kazakhstan, from 1966 to 1996.GeoResJ, 6, 141–163. https://doi.org/10.1016/j.grj.2015.02.014