Skip to main content
Dryad logo

Modeling and Monitoring Submerged Prehistoric Sites during Offshore Sand Dredging and Implications for the Study of Early Holocene Coastal Occupation of Southern California

Citation

Hildebrand, John; York, Andrew (2022), Modeling and Monitoring Submerged Prehistoric Sites during Offshore Sand Dredging and Implications for the Study of Early Holocene Coastal Occupation of Southern California, Dryad, Dataset, https://doi.org/10.6076/D1JP49

Abstract

Beach sand dredging projects off the coast of San Diego County in Southern California provide data for improved understanding of the strategraphic setting for early Holocene sediments and the potential for offshore presence of buried archaeological materials.  Geophysical data, core sediments, and analysis of recovered fossils allowed models to be developed for six offshore borrow sites within drown river valleys.  These site-specific models were tested during dredging operations, and the dredge spoil was monitored for archaeological materials.  Two of the borrow sites yielded stone bowls consistent with those found in previous offshore archaeological investigations in this region.  These artifacts, however, were determined to come from nearshore and lagoonal sediments, not appropriate for occupation, raising questions about both the function of stone bowls and the process that resulted in their deposition.  This project illustrates the potential for commercial development projects to yield information on offshore archaeological resources, as well as the challenges.

Methods

San Diego Association of Governments, Beach Replenishment Project

1999 Field Methods:

During January 1999, Sea Surveyor, Inc. conducted marine geophysical surveys and vibracore investigations along the San Diego coastline. The geophysical surveys of borrow sites were conducted using the 48' survey boat "WESTWIND". A differential GPS antenna was installed on the mast of the vessel, directly over the transducer well. The subbottom profiling included two acoustic sources: a Datasonics chirp II sonar, operating a frequency of 3-16 kHz, and a Applied Acoustic Engineering Geopulse system operating at 800 -2000 Hz.  The Geopulse acoustic source and hydrophones were towed alongside the vessel and 17' behind the GPS navigation antenna.   The seismic reflection data from both systems were displayed on an EPC Model 1086 thermal graphic recorder. No additional data processing was conducted on the graphic recorder output.

The 165' vessel "AMERICAN PATRIOT" and 20' ALPINE vibratory corer were used to collect sediment core samples at 125 locations in 10 proposed sand borrow sites during 18-24 January 1999. The ALPINE Vibracorer uses 4”-diameter steel barrels for collecting the core samples, and 3.5” diameter cores in clear cellulose acetate butyrate (CAB) liners, which can be split laterally or longitudinally for inspection and geotechnical logging and subsampling. After the vibracore collected a sediment sample, the sediment cores were extracted from the vibracore barrel, and the liner was cut at 5' intervals, capped, taped, and labeled. The collected cores were then shuttled to Oceanside Harbor, where geotechnical personnel split, inspected, described, logged, and subsampled the core samples for laboratory analyses. Subsamples from the sediment cores were transported to MEC Analytical Systems' laboratory in Carlsbad, California for grain-size and chemical analyses.

After the sediment cores were split longitudinally, logged, and photographed, the geotechnical personnel collected one or more representative subsamples from the top 1-3 layers of the core. The sediment subsamples were sealed within transparent ziplock bags, labeled, and transported to MEC Analytical Systems' Carlsbad laboratory for grain-size analyses. A total of 235 sediment samples were analyzed for grain-size. In the laboratory, approximately 40 grams of sediment from each subsample was weighed into a coors dish and placed into an oven to dry overnight. Once dry, the sample weight was determined by first weighing the sample in the coors dish, followed by weighing the dish after the sediment sample had been placed in the sieves. The sieves were stacked on top of one another with the sieve having the largest screen mesh diameter above sieves having progressively smaller screen mesh diameters. The sample was shaken for 10-minutes to sieve the sediment, which left the coarsest material on the upper screens and allowed the finer particles to fall through to the bottom. Once shaken, the contents of each sieve were weighed and the results were entered onto a data sheet. The data sheet was then entered into a computer spreadsheet, and the percentage (by weight) of gravels, sands, and silts were calculated.

More information on the 1999 field operations can be found at: https://www.sandag.org/index.asp?publicationid=590&fuseaction=publications.detail

2008 Field Methods:

The seismic reflection survey was initiated on October 6 and 7, 2008 using Fugro Inc. geophysical systems on the vessel "Julie Ann", a 24-foot aluminum hull vessel. The Starfix Seis navigation system was interfaced to a Trimble 12-channel GPS receiver with an integrated Starfix Differential receiver. The Trimble unit receives ranging information from the same satellites as the Starfix differential reference stations. These corrections are applied to the DGPS receiver's satellite data to produce an accurate (±5 feet) position of the vessel in real-time and a post processed accuracy of less than 2.5 feet. The differentially-corrected position from the Trimble receiver is then passed to the navigation computer.

The seismic reflection system used a mechanical "boomer" energy source and a multichannel, Geo-Eel hydrophone array. An Applied Acoustics Engineering AA300J portable seismic energy source was used to power the AA300 boomer plate towed from a catamaran configuration. The boomer plate is an electro- mechanical transducer made of an insulated metal plate and a rubber diaphragm adjacent to a flat wound electrical coil. A short duration high-energy pulse is discharged from the AA300J energy source into the coil and the resulting magnetic field repels the metal plate in the transducer. The plate motion is transferred to the water by the rubber diaphragm, generating a broad-band acoustic pulse that does not have strong cavitations or ringing. The Applied Acoustics Model AA300 sub-bottom tow fish was deployed and towed from the starboard side stern of the vessel. Sufficient tow cable was deployed such that the tow fish was clear of and beneath the vessel's wake. The system was triggered at an 8 to 10 Hz pulse rate and swept frequency range between 2 to 10 kHz. The recorded record length was dependent on water depth. Navigation fix marks were sent to the systems' printer every 100 meters down the survey line. All navigation information and sub-bottom data were time tagged and logged to a hard drive.

The reflected acoustic pulse, generated by the boomer source, was received by a multichannel Geo-Eel hydrophone array. The Geo-Eel includes hydrophones enclosed in a silica gel at specified interval. All track lines used a 24-channel (location dependent) hydrophone array. The hydrophone array included 5.1 foot (1.56 meter) group spacing of the channels. The raw data recorded by the hydrophone channels was logged to a Geometrics CNT recording system in a SEG-Y format for later post-processing to an accuracy of within 2.5% of penetration depth.

Seismic data processing was conducted by Mark Legg (Legg Geophysical), and the following description of the processing steps applied to the 2008 data was provided by him.

"Data processing of the multichannel seismic reflection profiles acquired offshore San Diego County, followed the basic premise of the wavelet processing method used in the petroleum exploration industry.  A single-plate boomer source type was used combined with a 24-channel hydrophone streamer configuration, with the objective high resolution at shallow sub-bottom depths (<100 m).  The boomer produced a source energy level of about 500 Joules over a broad bandwidth ranging from about 100 Hz to more than 1,000 Hz.  The 24-channel mini-streamer had a group interval of 1.56 m and a near trace to source offset of 3.125 m.  The short offset and group interval of the mini-streamer provides good imaging of the seafloor and shallow sub-bottom, while the far offset is sufficient to provide some water bottom multiple suppression in shallow water areas.  The short hydrophone group intervals of the streamer are very important for avoiding spatial aliasing of these digitally sampled data, thereby improving the overall signal-to-noise ratio of the seismic data.

Basic seismic data processing consists of filtering in both time and space, deconvolution to provide a sharper and more consistent seismic wavelet for interpretation, correction of normal moveout due to varying subsurface velocity structure and source-to-hydrophone offset, stacking of data traces to increase the signal-to-noise ratio, and migration to put the reflecting horizons back into their proper lateral positions.  A more detailed discussion of these steps, and the specific parameters for the processing flows for mini-streamer data follows.

Initially, the raw segy data must be loaded into the seismic data processing workstation.  With the simple and regular geometry of continuous marine seismic reflection profiling, where the offset from source to receiver positions is constant along the profile, these offset values are loaded into the trace headers during the data loading phase.  In order to retrieve the original field record numbers after data are stacked, a simple definition of shot points or station numbering scheme is applied with the first shot point defined as #101, and subsequent shot points numbered sequentially based on the particular recording geometry.  For this project, the shot point interval was nominally 0.78 m, which is equal to the common-mid-point spacing and equal to one-half the hydrophone group interval (1.56 m).  This geometry provides subsurface coverage at 2400 percent, i.e., nominal fold equal to 24. The station numbering scheme is designed to match the CMP numbering scheme, so the working shot point numbers during processing will have an increment of one (1) for this geometry.

The first step used in the processing flow, after loading the segy data into the processing workstation, is to filter and scale the data.  Preliminary review of data traces to edit or “kill” bad traces is performed.  This was particularly important for data acquired during the more severe sea state, with wind waves and short period swell that creates frequent “noise bursts” in the streamer.  An anti-alias filter is used as a band-pass filter to avoid aliasing in the time domain above the Nyquist frequency (2,000 Hz for 0.25-ms sample rate) and to remove low-frequency streamer noise, like bulge waves and water wave motion.  An Ormsby filter (trapezoid band-pass shape) was used with parameters of 10/15-1500/1800 Hz for the 24-channel mini-streamer data.  Scaling was then applied in two parts: first to remove the geometrical spreading attenuation with a time varying exponential function with an exponent of 0.3, and second to equalize the average amplitudes of each trace in the data set, using an RMS scaling factor for a window of data with reasonable signal-to-noise ratio, i.e., below the direct water wave and strong water bottom reflection and above areas where data consist mostly of background noise.

Spiking deconvolution is applied to shrink the original source wavelet down to an “ideal” zero-phase wavelet that is consistent from trace-to-trace and record to record.  With infinite bandwidth, this ideal trace would be a delta function, or spike at the appropriate arrival time.  For real band-limited data, a Ricker or similar symmetrical wavelet with minimal side-lobes is desired.  For these high-resolution data, spiking deconvolution used an operator length of 20-ms, and was designed using a window below the water bottom reflection where good signal-to-noise ratio is observed.  Another Ormsby filter is applied after deconvolution to eliminate high-frequency noise, using the parameters of 10/15-1500/1800 Hz to maintain broad bandwidth and eliminate some low frequency streamer noise.

After the first trace processing and editing steps, frequency-wavenumber (FK) analysis is done on select shot records to design FK filters to attack spatial aliasing.  Aliasing due to inadequate sampling in the spatial domain is often overlooked and may result in data artifacts from aliased high-frequency events that may appear as real reflection events.  For marine data, where the velocity of sound in water is about 1500 m/s, we can predict the frequencies where coherent noise traveling through the water past the streamer may become aliased: for the 1.56-mgi streamer, the spatial Nyquist frequency is 480 Hz.  These frequencies are lower than much of the source energy, and so array forming in the streamer must be accomplished to attenuate noise traveling horizontally in the water column.  Direct source to streamer wave propagation produces this coherent noise energy as does propellor noise in the water, from the shooting vessel as well as from other boats passing through the area.  An FK filter is well-suited to this, as it can preserve more of the vertically-incident high-frequency reflection signal.  The FK filters were designed to attenuate the low-velocity energy, assumed to be noise in the water column, and preserve high-velocity energy from subsurface reflections.  For this project, the FK filtering allowed us to expand the frequency range by attenuating the high-frequency source generated noise that travels horizontally along the streamer.  An Ormsby frequency domain filter follows the spatial (FK) filter to remove the higher frequency aliased noise; parameters used were 25/50 to 640/960 Hz.

From previous experience with the 24-channel mini-streamer, we observed that water-saturated sediment velocities in the shallow sub-bottom is very close to that of the water column, and sometimes slower due to gas content. For the mini-streamer data, the constant velocity brute stacks using the acoustic velocity of water at 5000 ft/sec were almost indistinguishable from the stacks made using the stacking velocity derived from longer offset 48-channel streamer data.  Therefore, all data were stacked using a normal moveout correction based on the constant 5000 ft/sec acoustic velocity.

The deconvolved and filtered traces were sorted into the Common Mid-Point (CMP) order, spatially filtered (FK filter), frequency filtered (Ormsby), normal-moveout corrected, and stacked to produce a CMP stack record section.  A static shift to correct for elevation (depth) of streamer and source, as well as the tidal elevation was also applied to the segy output data for the stack, and prior to migration for both plot and segy output.  [Note: the pdf plots of the stack data do not include the tidal or streamer and source elevation static correction.] The normal-moveout correction included a 75% stretch mute, to remove near surface data from the far offset hydrophone traces that get overly stretched in shallow water.  This stretch mute helps to maintain a better image of the water bottom and very shallow subsurface, which is stretched and smeared by the longer offset data.  Thus, stacking includes only the near offset traces for the water bottom and very shallow subsurface, but all traces for the deeper data.

For noisy data, acquired during the inclement weather, a 90th percentile “Alpha Trim” stack was applied to reduce noise bursts from water waves and cable jerks.  The Alpha Trim stack sums the trace values that fall within the 90th percentile of the median, thereby ignoring data spikes and outliers.  In general, the full stack with 1/N weighting when applied produced similar results, after trace editing had removed the most severe noise spikes in the raw data.  For the mini-streamer, nominal fold used in the stack is 2400%, i.e., 24 traces per CMP gather were summed and output at the CMP trace spacing of 0.78 m.  The high fold helps to minimize the effects of remaining noise spikes in the data.  With the 90% Alpha Trim stack, the actual fold varies for each trace, but is generally around 21-22.  Navigation and tidal static corrections were written to the trace headers of the stacked data, which then were filtered and output in segy format for loading into the interpretation workstation.  Ormsby filter parameters used after stack were: 25/50-640/960 Hz for the mini-streamer data.

Stacked data contain hyperbolic reflections and diffractions that need to be collapsed into proper spatial locations to further sharpen image of subsurface reflection horizons and faults.  We used a simple frequency-wavenumber migration, which works for all dip angles.  Because these high-resolution data involve shallow, mostly water-saturated subsurface sediments, a constant velocity of 4800 ft/sec was used for the migration.  Migration velocities are generally lower than stacking velocities because the latter are affected by horizon dip, and migration depends on the actual velocity of the geologic layers, i.e., interval velocities.  Data were filtered after migration and saved as segy data for loading into the interpretation workstation.  Ormsby filter parameters were: 90/120-640/960 Hz for the mini-streamer data.  The high-cut for these filters is designed to avoid spatial aliasing and remove high-frequency noise from the image.  The low-cut for these filters is designed to provide a higher resolution image for the mini-streamer and to maintain at least two octaves bandwidth.  A “top mute” was applied to the stack data prior to migration, to eliminate stack noise in the water column, which could “wrap around” to the bottom of the migrated traces during the FK migration.  These migrated noise spikes would appear as migration “smiles” that can obscure the primary reflection signal and complicate interpretation of the data. In addition to the segy output files, pdf plots of both stack and migrated data were prepared at a constant vertical exaggeration of about 10:1, for a seismic velocity of 1500 m/sec."

The M/V Supplier, a 63-foot steel work boat built as a military landing craft, was used as the platform for Vibracore operations between October 29 and November 8, 2008. A model 271 B Alpine Pneumatic Vibracore, configured to take cores 20 feet in length, was used, the same as the unit used for the 1999 investigations.

When the vibracore reached penetrations of up to about 18 feet, the unit was then retrieved to the support vessel, where the sediment contained in the plastic liner was extracted from the core barrel. While on board ship, the plastic liner was cut into 5-foot long sections for ease of handling, and to allow preliminary geologic logging of the collected sediment. More detailed geologic logging (requiring cutting the plastic liner along its length, then splitting the sediment core) was done onshore in a geotechnical laboratory.  Geotechnical logging of the recovered material included descriptions of the recovered sediment using the Unified Soil Classification System, and a Munsell soil color chart. Color digital photographs were taken of the split sediment samples. The color descriptions and photographs were of the sediment in a wet condition.

More information on the 2008 field operations can be found at: https://www.sandag.org/uploads/projectid/projectid_358_12897.pdf

Usage Notes

Track lines plots for sub bottom profiling data are represented by files with names:

TracklineDATE.pdf  where DATE is either 1999 or 2008.

These files use a Lambert Conformal Conic projection with the North American Datum 1983, California Zone 6 in feet.  The location of individual seismic lines are designated with a number at the beginning of the line, and distance along the track is designated in meters by the location of the CMP (common-mid-point) stack of the seismic data.  

Seismic data files are represented by files with names:

SITE_LINE#-DATE.pdf where SITE refers to the survey area, LINE# is the line number within that area, DATE is either 1999 or 2008. Files from 1999 are scans from the original graphic recorder output with no additional processing. They show the 3.5 kHz chirp sonar record on top and the Geopulse record on the bottom. The locations of vibracores are shown as annotations (blue triangles with core number). The vertical depth is given at 5 m increments. Files from 2008 were digitally recorded, and are the result of stacking and FK migration using a 24 channel streamer, as described in the methods section. Distance along track is given in m, and depth is given as time (ms) which can be nominally converted to depth using a two-way travel time and assumption of ~1500 m/s (0.75 m/ms), so that major time marks (10 ms) give 7.5 m depth increments. 

Vibracore Logs are represented by files with names:

VibracoreDATE.pdf where DATE is the year of the data collection. These plots give the sediment classification as a function of depth within each core.  Locations of grain size measurements are indicated at the appropriate depth. For the 2008 cores, field photographs are included at the end of the files.

GrainSize1999.pdf gives grain size distribution plots for samples obtained from cores collected in 1999. 

Funding

San Diego Association of Governments