Multichannel seismic-reflection and navigation data collected using SIG ELC1200 and Applied Acoustics Delta sparkers and Geometrics GeoEel digital streamers during USGS field activity 2020-014-FA, Southwest of Puerto Rico, March 2020

PROCESS STEP 1:
Shearwater Reveal (version 4.1) and Kingdom Suite (version 2017) seismic processing software was used to execute the following processing flows to produce SEG-Y files of profiles processed through post-stack time migration and deconvolution.
1. Import SEG-D sequences: SegDRead read raw Geometrics SEG-D shot sequence files, extracted navigation fixes from the external headers, and wrote them to new header words, and converted the source lat/lon positions from seconds of arc to decimal degrees (NRP_LAT, NRP_LON). UTMLatLong projected the geographic navigation to UTM Zone 19N WGS 84 meters (NRP_X, and NRP_Y). Output wrote the trace sequence to Reveal formatted ".seis" files.
2. Layback geometry assignment, delay removal, and trace edits: Input read sequence files sorted by FFID/CHANNEL. The Python module ShotlineLayback (developed by Nathan Miller of USGS-WHCMSC) defined the source and streamer geometry based on measured horizontal offsets from the navigation reference point (NRP) to the center of the sources (cos) and the centers of the first and last 1.5625 m spaced channel groups. The following lists the linear offsets from the NRP to center of source (COS) and index channels (IC) used to define layback geometry.
NRP to COS, NRP to IC offset(IC#) -31.13, -52.53(1)/-101.9(32) - Line0003 - Line0004a and Line0008c -38.93, -52.53(1)/-101.9(32) - Line0005 - Line0007a, Line0011 - Line0013b, and Line0015b - Line0017m -36.13, -52.53(1)/-101.9(32) - Line0009, Line0010, and Line0014
The module interpolated a sail line from the shot NRP positions (NRP_X and NRP_Y), then computed layback positions for the source and channel groups (values were interpolated for all channels in between the defined index channels) for each FFID by translating them back along the sail line by their respective offsets. Layback midpoint positions along the sail line were also computed for each shot/receiver pair. HeaderMath computed and populated a new header word for trace offset using OFFSET = ABS(NRP2CHAN) - ABS(NRP2COS). CMP bin spacing was set to either 0.78125 m (for shallow water lines; Line003, Line004, Line 004a, Line 8c, Line 9, and Line 10) or 3.125 m, the first cmp to 1, and CMP coordinates (BIN_X and BIN_Y) were computed. DBWrite wrote the layback geometry source positions (in UTM 19N WGS84 meters), FFID number, year, day number, and UTC times for the shots in each line to ASCII CSV files.
ApplyStatic shifted traces (except for Line0005, Line0008c, Line0009, Line0010, and Line0014) by -50 milliseconds to account for the recording delay used during acquisition and adjusted the trace start and end times accordingly. TracesEdits removed channels 7, 15, and 31 from output gathers due to high noise levels.
4. Apply static corrections:
For deep water lines (all except Line003, Line004, Line0004a, Line0008c, Line0009, and Line0010): Readings from RBRSolo-D pressure depth recorders spaced along the streamer were inserted into the trace headers for all lines. Input read the geometry corrected trace files. Text files prepared to contain UTC time, offset location of the pressure depth probe along the streamer, and pressure depth values in meters were used as inputs for DBMerge, which matched times and trace offsets in the trace headers with those in the input text files (through interpolation) and populated a new receiver depth header word with pressure depth values for each channel of the active section. HeaderMath also populated a new header word containing an estimated source depth of 1 m, then converted the sum of both static values from meters to milliseconds (dividing by an assumed water column sound speed velocity of 1.5 m/ms) and wrote it to a new header word. The source/receiver static value was then applied to the traces. Static corrections using the depth loggers successfully compensates for vertical movement of the streamer. Output wrote the static corrected traces to new files.
For shallow water lines (Line0003, Line0004, Line0004a, Line0008c, Line0009, and Line0010): Applying pressure depth static corrections to the shallow water did not produce satisfactory results, most likely due to horizontal movement of the streamer. Consequently, an alternate method to flatten both horizontal and vertical movement was applied. A Reveal flow used Output to write single channel SEG-Y format files for channels 1 to 32 for each line. The single channel (common offset) SEG-Y files were imported into Kingdom Suite (version 2017). The 2D-Hunt auto picker was used to digitize the seafloor reflection and direct arrival on all lines for traces where the two events did not interfere with one another. This resulted in pick on channels 1-16 for all lines except for Line0003 and Line0008c, in which picks were made for channels 1-12 and 1-8, respectively. Except for Line0004 and Line0009, a seven-trace moving average filter was applied to all seafloor horizon picks. An ASCII CSV file of the unfiltered seafloor, filtered seafloor, and direct arrival horizons was exported for each line and channel in which the horizons were digitized containing LineName (including channel number eg."_1"), FFID, and horizon two-way travel time values. All files were edited in the VI editor to eliminate any text other than the channel name from the LineName, and the LineName header was renamed to channel. Direct arrival CSV files were imported into Excel (version 16.48) where one-way direct arrival travel times, average one-way travel direct arrival times, and the difference between the digitized and average one-way direct arrival travel times (as a new field) were calculated for each line-channel. In Reveal, Input read the geometry corrected trace files. The CSV text files containing the unfiltered and filtered seafloor horizons were used as inputs for two DBMerge modules, which matched FFIDs and trace offsets in the trace headers with those in the input text files and populated new unfiltered and filtered seafloor header values. HeaderMath populated an additional header word with the calculated difference between the filtered and unfiltered seafloor horizons ((unfiltered seafloor - filtered seafloor) * -1), and ApplyStatic shifted traces by the resulting values. For all lines (Line003, Line004, Line0004a, Line0008c, Line0009, and Line0010) the CSV text files containing the difference between digitized and average one-way direct arrival travel times were used as inputs to an additional DBMerge module, which matched FFIDs and trace offsets in the trace headers with those in the input text files and populated a new difference direct arrival time header value, HeaderMath converted the difference direct arrival time back to two-way travel time, and ApplyStatic shifted traces by the resulting values. Output wrote the static corrected traces to new files.
5. Trace preprocessing and noise reduction: Import read files with geometry and static corrections. For Line0004, Line0004a, and Line0009, Despike estimated local median RMS amplitudes within 10 ms windows (specifying 11 traces per window, 25% trace and time window overlap, nearest neighbor interpolation, and replacement scalar of 2), then used the median RMS amplitudes to replace samples that exceeded them by 5 times. The remaining processes were applied to all lines. FXDecon was applied (specifying 5 points, 11 traces per window, 100 ms window length, minimum frequency of 30 or 200 Hz for Delta and SIG sources, respectively, maximum frequency of 1000 or 1400 Hz for Delta and SIG sources, respectively, and 1 percent prewhitening) to attenuate random noise most likely resulting from rough sea states during acquisition. A polygonal FKFilter designed in shot gather f-k spectra for each line was used to rejected noise energy outside of the primary reflections. Trace were Bandpass filtered (300-400-1300-1400 Hz for SIG ELC1200 lines or 70–90–900–1000 Hz for Applied Acoustics Delta lines) and Output wrote preprocessed, noise-reduced traces to new files.
6. Constant velocity stack, sea floor picking, and 1D velocity model creation: To produce a quality control stack, Input read the preprocessed, noise-reduced, and bandpass filtered traces. NormalMoveout converted traces to zero offset assuming a constant 1500 m/s sound speed through water, Stack produced a single trace per CMP bin, and Output wrote the stacked traces to new files. Sea floor reflection times were digitized for each line and saved to new horizon tables. 1D velocity models were created for each line using VelocityLayer (a python module developed for Reveal by Nathan Miller, USGS), to set RMS stacking velocities of 1500 and 2000 m/s above and below the digitized seafloor horizon, respectively. Trace2Table wrote the 1D velocity models to new database tables.
7. 1D velocity model stack and post-stack deconvolution: Input read the preprocessed, noise-reduced, and bandpass filtered traces, NormalMoveout converted traces to zero offset using the 1D velocity model table, Stack produced a single trace per CMP bin, and Output wrote the 1D velocity model stacked traces to new files.
8. Phase shift post-stack migration and migrated stack seafloor picking: Input read the 1D velocity model stacked traces, PostStackMigration performed a phase-shift time migration using the 1D velocity model (specifying maximum frequencies of 1000 or 1400 Hz for Delta and SIG sources, respectively, and the same bin spacing as the input (0.78125 or 3.125 m), and Output wrote the migrated stacked traces to new files. Subsequent to migration, the predicted sea floor picks were overlaid on the migrated trace display, manually adjusted to more closely approximate the migrated sea floor, and saved to new files.
9. Post-stack deconvolution, time-dependent gain, and top mute: Input wrote read the migrated stacked traces, Deconvolution applied a spiking deconvolution filter (designed using a 2 ms gap length, 50 ms operator length, and 0.2 percent prewhitening factor), Table2Header wrote the migration adjusted sea floor picks to a water bottom time header word, TraceMath applied a time dependent gain function to the traces relative to the migration adjusted digitized seafloor horizon (ifelse(TIME > WBTIME1 - 300, sample * pow((TIME-(WBTIME1-300))/1000,1), 0)), Mute applied a top mute above the migration adjusted digitized seafloor horizon, and output wrote the traces to new files.
10. SEG-Y output: Input read the migrated stacked deconvolved files. UTMLatLong projected the CMP bin positions from UTM 19N WGS84 meters to geographic and wrote them to new header words. HeaderMath converted the geographic coordinates from decimal degrees to seconds of arc multiplied by a scaler of 100, and set the coordinate unit header to 2 accordingly. DBWrite wrote the CMP positions (in UTM 18N WGS84 meters), CMP number, and year header words to ASCII CSV text files by line. Output wrote the stacked migrated deconvolved traces to SEG-Y Rev. 1 format (32-Bit IEEE floating point). Each output SEG-Y file contains a textural file header similar to the example from Line0003 included below.
Example migrated stacked file SEG-Y Textural File Header:
C 1 U.S. GEOLOGICAL SURVEY COASTAL AND MARINE HAZARDS AND RESOURCES PROGRAM C 2 SURVEY_ID: 2020-014-FA AREA: ATLANTIC OCEAN, CARIBBEAN SEA, PUERTO RICO, C 3 VESSEL: R/V SULTANA, YEAR: 2020 LINENAME: LINE0003 C 4 C 5 ACQUISITION: SIG MINI SPARKER, 50 M, 32 CHANNEL GEOEEL C 6 HYDROPHONE STREAMER (1.5625 M GROUPS), 0.25 MS RECORDING SAMPLE INTERVAL, C 7 0.5 SEC RECORD LENGTH, RECORDED IN SEG-D FORMAT. C 8 C 9 C10 PROCESSING: IMPORT SEG-D SEQUENCE FILES, EXTRACT NAVIGATION FIXES FROM C11 EXTERNAL HEADERS, MERGE SEQUENCE FILES TO SEIS FORMAT FILE, DEFINE C12 GEOMETRY WITH CMP BINS (SPACED 0.78125-M) USING NRP NAVIGATION AND C13 LAYBACK OFFSETS, REMOVE DELAY AND SET TRACE LENGTH, REMOVE C14 CHANNELS 7 AND 13-32, APPLY FREQUENCY DOMAIN BANDPASS FILTER C15 300-400-1300-1400 (Hz) APPLY SRC/REC STATIC CORRECTIONS FROM SWELL FILTER C16 AND FLATTEN TO DIFFERENCE BETWEEN DIRECT ARRIVAL AND MEAN DIRECT ARRIVAL C17 FOR THE LINE, APPLY NOISE FILTERS DESPIKE, FX DECON, AND FK FILTER, NORMAL C18 MOVEOUT WITH VELOCITIES 1500 M/S (WATER) AND 2000 M/S (SEDIMENT), STACK, C19 PHASE SHIFT TIME MIGRATION 1D VELOCITY MODEL, SPIKING DECONVOLUTION, C20 TIME DEPENDENT GAIN RELATIVE TO WATER BOTTOM. C21 C22 OUTPUT: 32-BIT IEEE FLOATING POINT SEG-Y C23 RECORD LENGTH: 0.5 SEC C24 C25 CDP COORDINATES ARE STORED IN GEOGRAPHIC ARCSECONDS (SCALED BY 100) C26 DIVIDE BY 360000 FOR DECIMAL DEGREES. C27 COORDINATE SCALAR IN BYTES 71-72 C28 C29 CDP-X AND CDP-Y IN BYTES 181-184 AND 185-188 C30 CDP NUMBER STORED IN BYTES 21-24 C31 C32 C33 FOR ADDITIONAL INFORMATION CONCERNING THIS DATASET, CONTACT: C34 WAYNE BALDWIN (508) 548-8700 x-2226 WBALDWIN@USGS.GOV C35 U. S. GEOLOGICAL SURVEY C36 384 WOODS HOLE RD., WOODS HOLE, MASSACHUSETTS 02543 C37 C38 C39 C40
Post-cruise processing was conducted by David Foster and Wayne Baldwin.

PROCESS STEP 2:
A python script (GEnavtoSQL.py) imported shot and CMP navigation from ASCII CSV files (produced in earlier processing steps), projected UTM 19N WGS84 coordinates to Geographic WGS84 using pyProj (version 3.1.0), and the navigation data into a SpatiaLite (version 4.3.0) enabled SQLite (version 3.3.0) database, creating three tables containing point geometries. The first contained records for all shots by line, the second contained records of all CMPs by line, and the third maintained records for the first and last CMPs, and CMPs at multiples of 500 by line. A 500-CMP interval was chosen because it corresponds to the annotation and tick interval provided along the top of the migrated, stacked profile images. The resulting database columns for the shot table consists of East, North (WGS84 UTM19N m), Lon, Lat (WGS84 dd), LineName, FFID, Year, JD_UTC (DDD:HH:MM:SS), SurveyID, VehicleID, and DeviceID. The resulting database columns for the CMP tables consist of East, North (WGS84 UTM18N m), Lon, Lat (WGS84 dd), LineName, ImageName, CMP, Year, SurveyID, VehicleID, and DeviceID. A third table was created to contain trackline geometries generated from the CMP point geometries for each line (sorted by LineName and CMP), and the line length in kilometers was calculated. The resulting database columns of the line geometry table consist of LineName, ImageName, CMP_init, CMP_end, SurveyID, VehicleID, DeviceID, and Length_km.

PROCESS STEP 3:
The shot, CMP, and 500 CMP points, and CMP tracklines were added into ArcGIS Pro (version 2.3.3) from the SQLite database (drag and drop from the database in Catalog pane folder connection), then exported (using the Feature Class to Feature Class geoprocessing tool) to the new Esri point and polyline shapefiles '2020-014-FA_MCS_shtnav.shp', '2020-014-FA_MCS_cmpnav.shp', '2020-014-FA_MCS_cmp500.shp', and '2020-014-FA_MCS_cmpTracklines.shp', respectively. The Export Feature Attribute to ASCII geoprocessing tool was used to produce ASCII CSV files from the attribute tables of '2020-014-FA_MCS_shtnav.shp' and '2020-014-FA_MCS_cmpnav.shp'.

The Seismic Unix (version 4.3) script 'plot_geomet' was used to read the migrated stacked deconvolved SEG-Y files and create variable density greyscale Postscript plots (using the Seismic Unix 'psimage' module) showing two-way travel time (seconds) along the y-axis (left margin) and shots along profile (labeled at 500 CMP intervals) on the x-axis (along top of profile). The Postscript images were then converted to 300 dpi PNG formatted images using ImageMagick convert (version 6.9.3-4). The output PNG files were named according to filename convention with '.psmig.png' indicating post-stack migration.