|
Wold-Crosby 1970 Seismic Survey of Flathead Lake Field Operations 2006 Data Recovery 2007-12 Data Archiving and Processing Dots-to-Data Images of the 1970 field operations were provided by Richard Wold and are used with permission. Images of the 2006 tape transcription were provided by Richard Hess and are used with permission. The captions for the following images are based on discussions with Richard Wold and various other people who are familiar with seismic surveying in lakes in the late 1960's and early 1970's. Steven and Paul Chelminski of Bolt Technology provided information on the air gun source. Peter Simpkin of IKB Technologies Limited provided information on the field chart recorder. Richard Hess of Vignettes Media in Toronto provided details on the field tape recorder and the instrumentation and methods employed for the 2006 recovery of data from an archive tape long stored in the US Geological Survey library at Woods Hole, MA. Ron Friedel provided input on field systems and operations. Handwritten notations on extant sections from the survey on file at the University of Montana Mansfield Library archives yielded some details. In addition to those mentioned above, a special note of gratitude is extended to Debbie Hutchinson of the USGS office at Woods Hole. Her fortuitous find of an old tape with some of the 1970 Flathead Lake seismic data made the processing discussed below possible. In the fall of 2015, members of the digital archive staff at the library of the Montana Bureau of Mines and Geology digitally enhanced the only known surviving line location and bathymetry map. The map is not on this site but is in the University of Montana digital archive. The work of Margaret Delaney and her staff is much appreciated. Feedback that could improve the captions is welcome. You can click on most of the images for a larger view. Use your browser's back button/arrow to return to this page. |
|||||
|
|
The survey was conducted in August of 1970 and used the University of Montana research station at Yellow Bay as a base of operations. This is an aerial view of Yellow Bay. The main facilities of the research station are in the window glare in the upper right of the image. According to Ron Friedel, at the same time the seismic survey was being conducted on the lake, John Sumner of the University of Arizona flew an aeromagnetic survey over the lake. |
||||
|
|
The principal investigators were Richard (Dick) Wold of the University of Wisconsin-Milwaukee (UW-M) and Gary Crosby of the University of Montana (UM). This image shows Dr. Crosby standing on the left. The person sitting cross-legged on the deck is believed to be Sidney Prahl, then a graduate student at UM. The man sitting on the bow is Ron Friedel, the UW-M technician who, along with Dick Wold, designed and constructed the seismic system. The student standing behind Ron is Carl Brzozowy, then a graduate student at UW-M. The name of the person standing between Crosby and Brzozowy is not known. Dick Wold, being the photographer, is not in any of the pictures of field operations. |
||||
|
|
Some members of the survey party had their families with them. |
||||
|
|
Dr. Crosby (left) and Ron Friedel appear to be working on the air gun source. The yellow object in the foreground is a float or pontoon from which the air gun was suspended to maintain a constant depth in the water. The suspension frame can be seen around the pontoon. This image suggests that the air gun was suspended about 2 ft below the water line. This is significant in recognizing the bubble pulse in the data. |
||||
|
|
This view of the stern of the survey boat shows two gasoline powered generators (yellow tops) and a gasoline powered air compressor (gray). A pressure gauge can be seen on the deck on a hose extending from the compressor. The dark rectangle in the bottom center of the image is the power conditioning unit seen in a later image. It is believed that the output from one generator was routed through the power conditioner to the tape recorder and the output from the other generator powered the chart recorder. |
||||
|
|
Here, Ron Friedel is working on the air gun. Ron believes that this is the 1 cu in gun. However, he notes that the survey had a 40 cu in gun that was used for some experiments. The records in the University of Montana Mansfield Library archives all seem to have been recorded with the 1 cu in gun because of the firing rate, approximately one shot every 4 s. Ron indicated that the larger gun had a firing rate of about 1 shot per minute because of the size of the compressor. Ron also points out the tube of silicone grease that is sitting on his knees. He said that he had to regularly disassemble the air gun and grease the O-rings in order to maintain proper operation. |
||||
|
|
Student working on one of the gas powered generators. The yellow object hanging on the boom over the stern of the boat is the pontoon that supported the air gun assembly. |
||||
|
|
This view shows the air gun pontoon that was towed. The air gun can be seen on the port side step at the end of the support harness. This image suggests that the air gun was no more than one meter below the surface of the water. This tends to reduce the effect of the bubble pulse. This also reduces the magnitude of the low frequency component of the air gun energy, which effectively reduces the depth of penetration of the seismic signal. |
||||
|
|
Here are the pontoon and air gun assembly in the water with the survey boat underway. View is to the northeast from the Bird Island area. |
||||
|
|
This is a view of the port side of the stern. This is believed to show the deployment of the hydrophone. Hydrophone cable is coiled on the roof of the boat's cabin and is secured with a rope tied to the hand rail. The cable extends to the red end of the boom and then just above the surface of the water to a white splash in the water. The amount of hydrophone cable that is deployed below the water surface is not known. One handwritten notation on a seismic section indicates the hydrophone was at 50 ft, but that notation does not indicate whether that distance was measured from the stern of the boat or from the deployed air gun. |
||||
|
|
The Wold 1982 USGS paper (MF 1433) states that boat positions were determined from paired theodolite readings from the shore. The exact process employed is not known. For example, were the two theodolite stations on opposite sides of the lake during a seismic traverse, or were both stations on the same side of the lake? Further, were the stations repositioned for each traverse? The Wold paper shows the positions of the lines on a basemap. However, only two lines are labeled, and not all of the surveyed lines are located on the published map, e.g., Lines A and B. The existing seismic sections in the University of Montana Mansfield Library archive show the times of “sextant” readings. No actual surveyor's notes are in the archive for either the theodolite or sextant readings. The sextant reading notations seem to be on the longer lines where the paired theodolite readings would have been ineffective because of low horizontal angle resolution or obstructions to sightings. The seismic sections have annotations, e.g., line name, survey date, clock times, and indications of the approximate locations of line ends, e.g., Blue Bay. A well-worn bathymetric map is in the seismic archive at the University of Montana. The paper map is a blue line copy of what was probably an original drawn on drafting film. It was prepared by Arnold Silverman, David Pevear, and Sidney Prahl, probably in 1971. The map was apparently never published. It shows the lines of Silverman's bathymetric surveys of Flathead Lake in the late 1960's and the lines of the Wold-Crosby seismic survey. The bathymetry and line locations in the Wold USGS paper appear to be copied from the Silverman et al. map. In addition to the physical map in the University of Montana archive, digital copies are available on-line, i.e., a scan of the surviving paper copy and the digitally enhanced version of the scan. |
||||
|
|
This appears to be a frequency regulated power supply. The label left of the meter can be read when the original scanned slide is enlarged. This unit probably stabilized the power that was used for the tape recorder. Stable power levels and alternating current frequency would have been more important for the tape recorder than for the paper recorder. |
||||
|
|
This is the tape recorder that was used during the survey. It was a readily available, economy, stereo recorder in the late 1960’s and early 1970’s. This particular model is a Magnecord 1024, which would have recorded two channels. As with the power conditioner, the label can be read in an enlarged image. The recorder used ¼ in tape on 7 in reels. This recorder, as arranged, appears to have two stereo amplifiers, which would imply four input/output channels. How these four channels were directed is not known. With no surviving field tapes, the actual recording scheme is not known. The second amplifier deck could have been, simply, a spare. Although the stock recorders usually ran at 3.75 and 7.5 in/s (abbreviated in 1970 as i.p.s.) some versions could be special-ordered that included an 1.88 (1-7/8) in/s speed. These recorders would have a low frequency response to 20 Hz or lower, especially at the slower speeds, but they did not go to DC. Low-frequency anomalies are always present in "direct" recording on magnetic tape as a result of head contour effects according to Richard Hess. This Magnecord is an AM audio recorder. The copy of the USGS archive tape that was digitized in 2006 was made as an FM recording on an instrumentation recorder, i.e., the archive tape is some sort of copy of the field tape(s). Ron Friedel says that an FM recorder was used at some point by the UW-M team because of its lower frequency response, but this image shows an AM recorder. The tape recorder was the heart of the seismograph system, which was designed and assembled at the University of Wisconsin-Milwaukee. Ron Friedel points out the two clocks mounted under the tape recorder. These were high precision, Swiss-made clocks that were used for timing the firing of the air gun. |
||||
|
|
This is a view of the paper chart recorder. The University of Montana has two copies of the sections for each line. One copy is clearly the field recording off of this device because of the hand written annotations. Comments by Ron Friedel and Peter Simpkin indicate that this is a Gifft wet paper recorder. The field recordings have aged to shades or orange and brown. Unfortunately, at some point, six or seven inches were cut off of the bottom of the field recordings. This leaves only 300 ms of data. The reason for trimming the sections is not known. Perhaps it was to make storage in a file folder easier. The second copy of the sections in the University archive is black dots on white paper. The 1982 USGS paper by Wold references post-acquisition processing by Sidney Prahl. The black-on-white sections were probably made on an EPC dry process recorder and may be the result of the referenced post-acquisition processing. These sections tend to show about 500 ms of data. The University of Montana archive groups the field recordings and the processed sections into separate items. Each item in the archive has a short description of the contents. |
||||
|
|
This is the same section as in the previous image just rotated into a more vertical perspective. The lake bottom reflection is clearly evident in this section. The base of the lake and basin fill sediments might be interpreted as the base of the generally featureless zone. In the water section, the bubble energy is obvious, and these oscillations are imprinted to some degree on the actual geologic signal. A bubble pulse reflection from the water bottom is obvious on the left side of the image approximately 25 ms after the primary water bottom event. (Remember, you can click on the image for a larger view. Use your back arrow to return here.) The timing lines are 25 ms apart. The delay between the primary and the first bubble pulse (25 ms) is consistent with the depth at which the air gun was apparently towed according to Paul Chelminski. Time zero on the section is not clear. The travel time to the lake floor reflection in the center of the section is either (approximately) 125 or 150 ms, depending on how time zero is defined, or between 300 and 375 ft, i.e., this section crosses a very deep part of the lake. This section is part of Line R, the east-most south-north line presented in the Wold 1982 paper. Line R line crosses Skidoo Bay and ends near Yellow Bay. The line was recorded on August 22, 1970. North is to the left in the image. This section is not in the data that were recovered from the USGS archive tape. Only 1/3 of the data from the survey was copied to the archive tape. |
||||
|
|
The original sections were made by “burning” a dot on the paper along the respective traces whenever the signal amplitude was in excess of some threshold value. Individual dots are discernible in the original sections. This is a portion of Line R. It is approximately the left half of the image above. One can see how the initially white to light tan paper has aged to orange or brown. The individual traces represent shots that are nominally 10 m apart. The timing lines are at 25 ms intervals. The gaps in the timing lines are at 5 minute intervals. Someone has drawn a pencil line to indicate the water bottom reflection. (The interpreted reflection and various annotations are more easily seen in an enlargement of the image. Just click the image on the left.) Also, at some point in time, the bottom of the original record was cut or torn off. A strip approximately 6 in wide has been removed from the bottom of the field records. The handwritten notations that are visible, but fuzzy, in the previous image have been transcribed from the removed strip onto the remaining part of the section. The black labels were added to the crop of the original image with a graphics editor. This section clearly shows the bubble event 25 ms below the primary waterbottom reflection. To the right of the arrow heads in this image, one can see another strong primary event and the bubble pulse 25 ms later. The bubble event was probably not recognized in the sections until the digitized data from the USGS archive tape were displayed in color. The full section of Line R is available through the University of Montana archive site. The file is 95 Mbytes. |
||||
|
The 2006 Data Recovery Project Field Operations 2007-12 Data Archiving and Processing Dots-to-Data |
|||||
|
Dick Wold left the University of Wisconsin-Milwaukee and then worked for the USGS for a number of years in the late 1970’s and early 1980’s. Gary Crosby left the University of Montana in 1973 and worked for a major petroleum company. When interest in the Wold-Crosby survey surfaced in early 2006, Bob Lankston learned that the USGS library in Woods Hole, MA, had an analog tape with Flathead Lake seismic data. Nancy Soderberg, the USGS archivist had very little information about the tape. However, she checked the tape out to Bob so that he could attempt to recover the analog data and convert them to a modern seismic format. In a modern format, the data could be deconvolved to sharpen the source signature and reduce multiples and migrated to collapse the diffraction arcs. Because the seismic system was built outside of the conventional seismic industry, no petroleum industry seismic lab in Houston, New Orleans, or Wales (UK) was able to read the tape. After considerable internet research, Bob found a service company in suburban Toronto that specializes in recovering data from old analog audio recordings, which the USGS archive tape seemed to be. Analysis of the archive tape and conversion of the analog signals to digital .wav files was done by Richard Hess of Vignettes Media. |
|||||
|
Because so little was known about the tape, the first thing that Richard Hess did was “develop” the tape. This is a process of applying a solution of “iron filings” to the tape and seeing how they align. This image shows the results of such a test on the USGS archive tape. Even before this test was done, however, it was clear that the archive tape was not a field tape. The Magnecord 1024 used ¼ in tape on 7 in reels. The archive tape was ½ in wide, as can be seen in this image, and on a 10 in reel. Therefore, the archive tape had to be a transcription of the original field tape or tapes. The number of tracks on the tape was not known. Various tape head standards exist for ½ in tape, including 2, 4, 7, 8, and more tracks. Richard determined that this tape was recorded with the IRIG format. IRIG (Inter-Range Instrumentation Group) was developed for missle-range work and is embodied in many versions of IRIG Standard 106. It is now considered obsolete. This tape was recorded on a machine with a two part head assembly with 7 interleaved tracks, i.e., 3 heads on one part of the assembly intercalated with 4 heads on the other part of the assembly. Fortunately, the data were on tracks 2 and 4, which were both on one of the two head assemblies and are indicated by the light gray stripes in the right half of the tape. While the field recorder was most likely a common audio deck, the transcribed archive tape was made with an instrumentation recorder, not an audio recorder. Audio and instrumentation are similar, but are not identical. Also, how the tape was wound on the reel was not known, i.e., did it need to be rewound to get to the beginning. Richard was able to resolve these and many other issues in order to read the tape in analog form and then to digitize the analog signals. |
|
||||
|
|
The USGS archive tape is on the right spindle on Richard Hess’ playback machine. While this Sony machine is essentially an audio recorder, Richard has a set of IRIG Standard 106 7-track heads specially mounted on a spare head block so that he can handle this obsolete format. The heads themselves were from a Racal Store 7DS instrumentation recorder that a colleague was coincidentally converting to an audio transport. Unfortunately, the Racal Store 7DS did not take the size reel that the Flathead Lake tape came on. The discovery of the tape and the analog to digital conversion benefited from a series of good luck events.
|
||||
|
|
This is a screen capture from Richard Hess’ early tape analysis efforts. The vertical red bars in the second track from the top, track 2 in the developed tape image above, are the high energy burst at the start of each seismic trace. They occur at approximately 4 s intervals. Other periodic energy bursts at 10-12 s intervals were interpreted to be residual noise from a previous tape erasure episode. |
||||
|
|
Direct playback and digitization of the analog tape yielded the upper trace. This did not have the appearance of seismic signals. Richard Hess did additional analysis of the digitized signals and research on the equipment of the 1970’s and deduced that the data on the USGS archive tape were FM. The lower trace is the result of demodulating the upper trace. This looks more like a seismic trace. The timing lines are 50 ms making this image a little more than 300 ms. |
||||
|
Once the details of the tape had been resolved, Richard Hess digitized the tape and made DVD's of the two tracks with seismic data. The files with 44 kHz sampling represent a digital image of the original recording. Richard decimated the 44 kHz files to 1 kHz and 8 kHz sampling for the digital seismic processing. The .wav files with 1kHz and 8kHz sampling are now in the University of Montana digital archive. |
|||||
|
In early 2007, Jenica Andersen, then an undergraduate at The University of Montana Department of Geosciences, digitally scanned all of the sections, handwritten notes, and interpretation sketches that had survived in the department for over 35 years. All of the scans are at 600 dots per inch. This makes the files for the field recordings very large. This scan resolution was chosen with the hope that subtle features like the hand written annotations on the sections would be preserved. The scanned sections and notes were posted to a digital archive at the university's Mansfield Library in early 2012. Also on the archive site are the original .wav files prepared by Richard Hess. The university archive is still in development. At some point, it will probably contain SEG-Y and SU format files of the data so that the user is freed from the burden of reformatting the .wav files. The archive also has an image of the Silverman, Pevear, and Prahl bathymetry and survey line map. Custody of the tape that was held by the USGS library at Woods Hole was transferred to the University of Montana Mansfield Library. After the .wav files were delivered by Richard Hess, Bill Menger, then manager of seismic processing at ConocoPhillips, set about converting the seismic data in the 1 kHz and 8 kHz .wav files to SEG-Y format using C code of his own design to read the .wav files and output SEG-Y files.. At the same time, Bob Lankston at Geoscience Integrations took a slightly different approach. Bob used an open source program called sox to convert the .wav file to tabular ASCII. The ASCII file was reformatted for input to Excel using gawk. Excel was chosen for extracting the individual traces because of its ready plotting capability. The files did not have any indication of trace starts, i.e., time zero. Therefore, a scheme had to be developed to recognize the high energy of the direct arrival at the hydrophone. A scheme was developed in Excel, and approximately 5000 traces were recovered from the file for one of the two tracks. The scheme was not perfect, and as much as 5% of the traces was not recovered. Once the algorithm was developed in Excel, it was ported over to C. Manual editing of the input data files resulted in all of the traces with usable data being recovered. With the C code, both seismic tracks were formatted into sets of 1 s traces. While Excel was satisfactory for plotting a few traces while developing the trace start scheme, it was not suitable for plotting large numbers of traces. The open source program gnuplot was used initially to plot the more or less 5000 traces. These sections were compared to the 1970 field recordings in order to decide which lines had been preserved on tape. The comparison showed that only about 1/3 of the data from the survey had been preserved on the USGS tape. In early 2007, the Seismic Unix (SU) package developed by the Center for Wave Phenomena (CWP) at the Colorado School of Mines was installed and has been used for processing and displaying the digital traces. The individual 1 s traces that were pulled from the .wav files were converted to .su format using the a2b function in SU. Once the traces were in SU format, headers were applied to each trace. Because the extracted traces were the result of a “guided guess” of the actual start time for each trace, a statics routine was applied in SU to improve the start times. This work was done on traces that were sampled at 8000 Hz in the analog to digital transcription. While the cross correlation-based statics routine aligned the direct arrivals, the true trace start time was not known. Because the sections showed strong waterbottom multiples, the primary to multiple time was used to set the time zero to waterbottom primary time. Once the trace starts were set satisfactorily, the stage was set for all manner of analysis of the data with SU. Homomorphic deconvolution to suppress the bubble pulse is among the most recent success stories. |
vv |
||||
|
|
|||||
|
The only lines that were recorded on the USGS archive tape were those surveyed in the northern part of Flathead Lake. Those were, from the north, Lines A, B, C, D, E, and F. Those lines represent only about 1/3 of the total survey. The data above are from Line F. Wold published Line F in his 1982 USGS poster (MF 1433). The set of sections plotted on this black-on-white medium is presumed to be the result of the post-acquisition processing that Wold credited to Sidney Prahl in the 1982 paper. The nature of that processing is not known. It was probably limited to band pass filtering and automatic gain control (agc). Timing lines in this image are at 100 ms intervals. This image is in the University of Montana digital archive. |
|||||
|
|
|||||
|
This is the same section (Line F) as in the previous image after digitization of the analog data on the archive tape that had been held by the USGS library at Woods Hole and processing with the Seismic Unix software developed by Colorado School of Mines. John Stockwell's help with SU, particularly the homomorphic deconvolution that was used to attenuate the bubble pulse, is gratefully acknowledged. The processing flow that produced this image is discussed in “Lankston, R. W., 2011, New display of the 1970 Flathead Lake seismic data: Northwest Geology, v. 40, p. 55-62”. This section shows only 400 ms while the image above of the 1970's processing shows 500 ms. Timing lines here are at 50 ms intervals. West is on the left. Shotpoints are approximately 10 m apart making this line about 9650 m long. Georeferencing of the Silverman et al. map in a GIS in 2013 indicated that Line F is closer to 11,200 m long, suggesting that the trace interval is slightly more than 10 m. |
|||||
|
2012-13 Dots-to-Data Experiment Field Operations 2006 Data Recovery 2007-12 Data Archiving and Processing |
|||||
|
The data that were available on the archive tape held by the USGS covered just six west-east lines in the northern half of the lake. These lines represent about 30% of the total survey. After parsing the .wav file into traces for use in Seismic Unix (SU) and application of various processing schemes as mentioned above, these six lines were exported in SEG-Y format for use in a seismic workstation. Unfortunately, the other 70% of the survey was not available for use in the workstation environment. This prompted a search on the internet for software that could convert the scanned images of the field or processed sections into digital data. Finding nothing, an idea came up that each column of dots and white space in the scanned image represented a time series of positive values, nominally 1 where a dot on the image appeared, and negative values, nominally -1 in white space. This time series was something like a seismic trace. A simple Python script was written to extract the series of 1's and -1's from each column, and each column was treated as a seismic trace. The quasi-trace was band pass filtered to provide a more oscillatory character. Of course, the true waveform characteristics were lost in the original paper sections. At this point, the data were in a digital form that could be imported to the workstation. |
|||||
|
|
|||||
|
The image above shows data from the eastern half, more or less, of line E. The left panel was prepared from the analog data digitized from the USGS archive tape. The panel on the left was generated from the scanned image of the processed section from the early 1970's work using the dots to data process. The image on the right shows the same geologic features as the one on the left. The colors of the events in the image on the right are more saturated because of the initial binary nature of the signals. |
|||||
|
top |
|||||
