Lee 000146

7/27/2019 Lee 000146

1/6

N e a r - s u r f a c e g e o p h y s i c s

146 The Leading Edge February 2011

SPECIAL SECTION: Ne a r - s ur fa c e g e op h y si c s

Subsurface characterization within urban centers iscritically important for city planners, municipalities,and engineers to estimate groundwater resources, trackcontaminants, assess earthquake or landslide hazards, andmany other similar objectives. Improving geophysical imagingmethods and results, while minimizing costs, provides greateropportunities for city/project planners and geophysicists aliketo take advantage of the improved characterization affordedby the particular method. Seismic reection results canprovide hydrogeologic constraints for groundwater models,provide slip rate estimates for active faults, or simply mapstratigraphy to provide target depth estimates. While manytraditional urban seismic transects have included the use ofvibroseis sources to improve reection signals and attenuatecultural noise, low-cost and high-quality near-surface seismic

reection data can be obtained within an urban environmentusing impulsive sources at a variety of scales and at productionrates that can signicantly exceed those of swept sources.

Sledgehammers and hydraulically powered acceleratedweight drops allow rapid acquisition rates through dense ur-ban corridors where the objective is to image targets in the up-per 1 km. In addition to the many extra steps often requiredto work in an urban area compared to more remote sites (e.g.,complex permitting process, public awareness/educationcampaigns, damage to/interruption of roads/utilities), cultur-ally noisy urban environments can provide additional chal-lenges to producing high-quality seismic reection results.

Urban recording can be complicated by problems unique tothis environment and can range from large-amplitude sig-nals from vehicles and machinery to electrical signals withinfrequency bands of interest for near-surface targets. Acquisi-tion methods designed to address both coherent and randomnoises include recording redundant, unstacked, unlteredeld records. Processing steps that improve data quality inthis setting include diversity stacking to attenuate large-am-plitude coherent (nonrepeatable) vehicle noise and subtrac-tion of power-line signals via match lters to retain reectionsignals near alternating current frequencies. Tese acquisitionand processing approaches allow for rapid and low-cost dataacquisition at the expense of moderately increased comput-

ing time and disk space. Here, I present acquisition and pro-cessing challenges that accompany urban seismic reectionsurveys with examples using low-cost seismic sources wherehigh-quality reection images have been obtained along busycity streets during peak traffic hours. Tese examples are fromearthquake hazards and groundwater studies and support thesuggestion that seismic reection imaging can provide a low-cost, high-quality component to many urban resource andhazard assessment studies.

Rapid and exible hammer seismic data

One of the challenges to acquiring quality seismic reection

data in an urban environment is the balance between an ide-

LEELIBERTY, Boise State University

alized prole for the geophysical target of interest and opti-

mal seismic prole location. City streets or urban corridorsmay not extend across the geologic targets of interest. Onebenet of small impulsive sources is an increased exibilityof survey location. Hitch-mounted weight drops or portablehammer sources allow surveys to be carried out along nar-row property easements (e.g., bike paths, road shoulders, pri-vate property) or on soft ground where heavy trucks are notpermitted (Figure 1). Operation of weight-drop sources mayrequire a traffic control plan similar to vibroseis acquisitionand a demonstration that the hammer will not impact citystreets or utilities. Generally, standard street constructiondesign will prevent damage from a large accelerated weightdrop source with an appropriately designed striker plate.

Compaction or damage of soft ground from a large ham-mer source and plate is generally not permanent. Becausehydraulical ly driven accelerated hammers are impulsive, shotgathers can be acquired and recorded in a few seconds. Incontrast, vibroseis records are generally acquired with longsweep and listen times that require tens of seconds to recorda single sweep. As a result of more rapid acquisition, either asignicantly greater number of shot gathers can be acquiredper source location and/or production rates can be greatlyimproved when comparing impulsive sources to swept sourc-es. In practice, 46 hammer shots per source location re-quire approximately one minute to complete, resulting in 60

source points per hour. Assuming cables and geophones can

Figure 1. (a) Brady Flinchum operating the Boise State Universityhitch-mounted accelerated hammer source along a residential street inReno, Nevada. A rubber mat is mounted below the plate to minimizeroad impact. (b) railer-mounted seismic source used to characterizegroundwater aquifer through a high-traffic area in Boise, Idaho. (c)Portable slide hammer seismic source used to connect two weight dropseismic proles along residential streets in Seattle, Washington.

Downloaded 31 Oct 2012 to 118.97.33.140. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

7/27/2019 Lee 000146

2/6February 2011 The Leading Edge 147


be deployed as rapidly as source production rates, eight hoursof shooting will yield 480 source positions per day. A typicalCDP survey with 5-m spacing will then yield a productionrate of nearly 2 km/d. Similar acquisition surveys using vi-broseis sources have yielded approximately 1 km/d for theauthor.

Unlike frequency-controlled vibroseis sources, hammersources can introduce larger-amplitude low-frequency sig-nals that may not be desired for reection imaging. Acqui-

sition using higher natural-frequency geophones can reducelow-frequency surface-wave recording while retaining thehigher-frequency signals. However, lower-frequency signalsmay dominate near-source channels and require either a sur-gical mute or lter to attenuate. Impulsive sources can pro-duce reected signals that are higher frequency than sweptsource capabilities (e.g., Bachrach and Nur, 1998); however,the variable amplitude with frequency response of impulsivesources often requires frequency balancing via shaping ltersor deconvolution to broaden the signal response.

Nonstationary noise

Vehicle noise is likely the largest-amplitude noise source en-

countered in an urban environment. Although a large truckcan saturate the dynamic range of modern seismographs, thetraffic noise is often lower frequency than the reection sig-nal produced from many high-resolution seismic reectionsurveys, and is nonstationary (Figure 2). Tis nonstationar-ity aspect is key to traffic noise attenuation. By recordingmany individual shots every few seconds, a standard vehicletraveling at residential speeds will move through the activerecording spread. For example, a vehicle moving at 65 km/

hr (18 m/s) will have moved 90 m in 5 s. Assuming 5 shotsper station at 5 s between shots, the vehicle will have moved450 m during the acquisition time for a given shot location.Tis distance is a typical spread length for a high-resolutionseismic survey. By recording individual shots in the eld,amplitude normalization or editing prior to stacking cansignicantly attenuate traffic noise. Generally, traffic will ap-pear as high-amplitude, low-frequency, coherent noise withsurface-wave velocities (Figure 2). With multiple shot recordsper station location to allow a vehicle to travel down the ac-tive spread, trace normalization prior to stacking will treatthe nonstationary traffic noise as a random signal that can be

attenuated with standard stacking procedures. A simple ap-

Figure 2. Shot gathers along a city street with four summed gathers: (a) vertical stack with no gain or lter showing vehicle noise at large offsets;(b) AGC prior to vertical stack with no lter; (c) AGC and band-pass lter; and (d) diversity stack (scaled by the inverse of the power) andband-pass lter. Note the traffic noise identied on (a) is attenuated on (b) and (c), but best removed on (d). Tese data were acquired along aresidential street in Seattle, Washington to identify and characterize active faults.


7/27/2019 Lee 000146

3/6148 The Leading Edge February 2011


proach is to scale each sample by the inverse of the power, ordiversity stack. Using this stacking approach, large-amplitudetraffic noise becomes a small-amplitude signal while repeat-

able reection signals will remain intact. Figure 2 comparesfour individual hammer shots from a trailer-mounted weightdrop (Figure 1) with signals to greater than 0.5 s and com-bined by: (a) a standard vertical-stack approach (sample aver-aging); (b) an amplitude gain control (AGC) prior to verticalstack with no band-pass lter; (c) AGC and band-pass lterprior to vertical stack; and (d) diversity stack and post-verti-cal-stack band-pass lter. Notice that the traffic noise identi-ed on (a) is attenuated on (b) and (c), but best removed on(d). Figure 2d supports the suggestion that the best resultsare obtained by using a diversity stack approach to reducingthe effects of large-amplitude nonstationary noise. In this ex-ample from Seattle, Washington, structures and stratigraphy

were clearly imaged in the upper 1 km at a rate of nearly 2km/d (e.g., Sherrod et al., 2008).

Attenuation of large-amplitude coherent noise in seismic

reection data has taken on new interest with simultaneoussource acquisition methods now employed in the industry toincrease production rates (e.g., Beasley, 2008). New process-ing methods can be directly applied to urban hammer seismicsurveys. High-fold hammer data with standard processingsteps can signicantly improve data quality while improvingproduction rates relative to vibroseis acquisition. Te abilityto acquire useful seismic data along the road shoulder of busycity streets comes at a cost of greater processing time and diskspace to handle the larger data set. However, the ability tooperate in urban centers, residential streets, and narrow ease-ments during daylight hours can minimize community dis-ruption and provide greater survey design exibility.

Figure 3. (a) Unltered shot gather pairs acquired along residential street in Seattle, Washington with large overhead power lines. Frequencyspectrum shows peak amplitudes at 60 and 180 Hz related to alternating current signals. (b) Notch lter response showing shot gathers removemuch of the power-line noise. Filter artifacts appear above rst arrivals and 60-Hz signals are removed. (c) Match lter design and subtractedfrom (a) to produce shot records where power-line signals are removed. Frequency spectrum is more consistent at 60 Hz where reection signalsremain in the shot record.


7/27/2019 Lee 000146

4/6150 The Leading Edge February 2011


Power-line noise

Power lines are difficult to avoid in ur-ban areas. Power-line signal strengthon a shot record can often be a func-tion of ground saturation and surfacematerials, and can overwhelm the sig-nal on a shot record (Figure 3). Gener-

ally, power-line noise appears as high-amplitude single-frequency signals (50or 60 Hz). Related higher harmonicfrequencies can often produce am-plitudes of similar magnitude withinthe frequency range of the desired re-ected signals. While notch lters willremove signals at a single frequency,this approach can have undesired ltereffects, including removal of reectedenergy within the notch band. Insteadof notch lters to remove these sig-

nals, signal subtraction by designing amatch lter is a more robust methodto remove this noise (e.g., Butler andRussell, 1993; Xia and Miller, 2000).Tis method allows reected signals toremain in the data set while removingthe high-amplitude narrow-frequency,cyclic response of the power-line noise.

One more consequence to ignoringpower-line signals or improperly re-moving power-line noise is the addeddifficulty of calculating a reliable resid-ual static model. Application of residualstatics with large-amplitude power-linesignals retained in the data often resultsin alignment of out-of-phase power-line signals within common midpointtraces. Removal of power-line signalsshould be performed on individual un-summed traces to ensure the greatestefficiency of the match lter removingpower-line signals.

Trigger timing errors

A contact closure or voltage-producing switch can trigger

a seismograph when using a hammer source. Tese triggersignals are usually telemetered to the seismic recorder viaground wire or radio. iming errors from contact closurescan result from dirt on the strike plate (from dragging theplate on the ground) while hammer switches can producespurious time breaks, variable pulse characteristics, and canvary sensitivity with increased usage. Radio triggers or trig-ger wires are sensitive to cultural noise interference (electricfences, power lines, etc.) that can also cause timing errorsupward of a wavelength or more. As a consequence of theinstability of time breaks, high-frequency data should notbe stacked without addressing and estimating timing errors.

Figure 4.Map, elevation prole, and migrated/depth-converted hammer seismic section fromReno, Nevada. Tis 1.4-km, 3-m source- and receiver-spaced prole was acquired in a singleday along a residential street. Te bedrock surface was constrained by borehole informationand seismic character. Reections overlying the bedrock surface typify alluvial stratigraphy thatdominates the Reno Basin.

Cross-correlation or residual static techniques can address

these timing errors. Residual statics approaches for near-surface seismic studies have been summarized by Pugin andPullan (2000).

Case studies

Reno, USA: active fault mapping. Five 120-channel hammerseismic proles were acquired along residential streets toidentify and characterize potentially active faults that crossthe downtown Reno, Nevada area (Liberty, 2010). Figure 4shows results from one prole that crosses a mapped fault atposition 3150 along Warren Way in south Reno. Tis 1.4-kmprole was acquired in a single day with a hitch-mounted


7/27/2019 Lee 000146



accelerated weight drop (Figure 1) using 3-m geophone andsource spacing. raffic proceeded along the residential road

while four individual shot records were recorded at 434 sta-tions. Te student crew placed geophones mostly in privateproperty easements along the road shoulders. Power lines ap-peared overhead and many geophone stations were skippeddue to road and driveway crossing. Processing steps included

cultural noise attenuation as described above and diversitystack to remove traffic signal.

A lateral change in basin stratigraphy and a relatively atbasal reector characterize this seismic image (Figure 4). Al-though I do not identify evidence for active normal fault-ing on the seismic prole, uvial architecture and bedrocktopography are mapped with the aid of nearby water wellsand clear seismic character. Te lack of reectors in the upper100 m of the northern portion of this prole suggests coarse-grained, laterally discontinuous alluvium dominates the basin

while alternating coarse and ne-grained laterally continuous

Figure 5. (top) Aerial photo and seismic reection prole location along a railroad andresidential street in Boise, Idaho. (bottom) Seismic reection image from a groundwater study inBoise, Idaho where prograding delta sands and mudstones dominate the stratigraphy in the upper1 km. Red dashed lines represent interpreted faults, the red circle represents a uvial channel andgroundwater target, and arrows represent the base of a shallow sand unit. wo deep wells act ascontrol points for depth conversion and geologic interpretation.

alluvial deposits are found along thesouthern portions of the prole. Rapiddeployment using a small weight dropprovided a clear image of subsurfacestratigraphy with a center frequency ofapproximately 100 Hz. Tis acquisi-tion approach has led to more seismic

proling at a greatly reduced cost com-pared to an equivalent vibroseis survey.

Boise, USA: aquifer characteriza-tion.Over the past decade, in collabo-ration with the state of Idaho, I haveacquired a number of seismic reec-tion proles throughout Boise, Idahoto map stratigraphy and geologicstructures with the intent of improv-ing the understanding of groundwaterresources for the residents of southwestIdaho (Liberty et al., 2001). Tese pro-

les have been acquired in downtownareas and residential neighborhoods.Figure 5 shows an aerial map and seis-mic prole from an area in Boise wherethe groundwater resource in the upper1 km was poorly understood due tocomplex basin stratigraphy, a low-per-meability mudstone-dominated basinsequence, faulting, and shallow in-terbedded basalt ows (e.g., Wood,1994).

I used a 120-channel seismographwith 5-m source and receiver stationspacing to produce a 60-fold seismicreection section (Figure 5). Portionsof this prole were acquired along arailroad right-of-way, while other por-tions were acquired along residential

streets. Vehicle traffic was encountered along the length ofthe prole and power lines were present along the length ofthe prole. Te engineered ll of a railroad grade can reducesource and receiver coupling in places and can produce sig-nicant near-surface lateral velocity contrasts. However, thesedifficulties did not prevent high-quality seismic reectiondata from being acquired along the length of the prole.

Te alternating mudstones and sands of the Idaho Groupprograding delta sequence overlie a shallow, yet extensivesand aquifer at 3050 m depth along the length of the prole(Figure 5; Wood, 1994; Liberty et al., 2010). Tis sand aqui-fer is characterized by a seismically transparent zone overly-ing west-dipping strata. Additionally, a shallow sand channelcuts the central portion of the seismic prole (position 2750)and may provide a good source for shallow groundwater. Off-set reectors at depth suggest normal faults have shaped theevolution of the basin and may inuence deep groundwaterow. Water wells, spaced more than 1 km apart that extend to


7/27/2019 Lee 000146



~300 m depth, show similar lithologies in each well. Howev-er, this seismic image suggests different stratigraphic units areencountered within each well due to depositional dips thatapproach 10. Terefore, groundwater pumped from the twoadjacent wells will draw from different aquifer units, and insome cases, these aquifers are separated by thick mudstone se-quences that inhibit lateral groundwater ow. Seismic imag-

ing in the Boise area has led to an increased understanding ofthe groundwater resource and the stratigraphy and structuresof the Western Snake River Plain.

Summary

High-quality seismic reection data can be obtained in ur-ban areas using low-cost hammer sources along city streetsand narrow urban corridors. Acquisition and processingstrategies can signicantly reduce much of the large-ampli-tude signal related to traffic and other sources of culturalnoise. Strategies such as diversity stacking common sourcepoint data will balance high-amplitude coherent traffic

signals relative to reection returns, but require recordingindividual eld shots. Processing steps that include matchlters to subtract power-line signals help retain a broad-bandreection response while removing high-amplitude, narrowbandwidth cyclic signals that can overwhelm reection am-plitudes. Although off-time (evening or weekend) acquisitionalong culturally quiet transects is optimal, these opportuni-ties may be unavailable or undesired in urban and suburbanareas. Examples from earthquake hazards and groundwaterstudies show high-quality hammer seismic reection resultsthrough urban corridors at a low cost. Tese inexpensive seis-mic sources, combined with acquisition and processing strat-egies described here, can lead to an increased understandingof many geologic and hydrogeologic targets at a variety ofscales to benet urban and project planners, reduce hazardrisks, or quantify urban resources.

ReferencesBachrach, R. and A. Nur, 1998, High-resolution shallow-seismic ex-

periments in sand, Part I: Water table, uid ow, and saturation:Geophysics, 63, no. 4, 12251233, doi:10.1190/1.1444423.

Beasley, C. J. 2008, A new look at marine simultaneous sources: TeLeading Edge, 27, no. 7, 914917, doi:10.1190/1.2954033.

Butler, K. E. and R. D. Russell, 1993, 993, Subtraction of power-line harmonics from geophysical records: Geophysics, 58, no. 6,898903, doi:10.1190/1.1443474.

Liberty, L. M., 2010, Seismic reection imaging of the Mount Rosefault zone, Reno, Nevada: Report to the U.S. Geological NationalEarthquake Hazards Reduction (NEHRP) program, CGISS ech-nical Report 1002, http://earthquake.usgs.gov/research/external/reports/G09AP00071.pdf

Liberty, L. M., S. H. Wood, and L. Barrash, 2001, Seismic reectionimaging of hydrostratigraphic facies in Boise: A tale of three scales:71st Annual International Meeting, SEG, Expanded Abstracts,13931396.

Liberty, L. M., 2010, Geophysical characterization at the North Adaand East Ada sites: A 2010 Idaho Department of Water Resourcesreport, CGISS echnical Report 1001.

Pugin, A. and S. E. Pullan, 2000, First-arrival alignment static cor-rections applied to shallow seismic reection data: Journal of En-vironmental and Engineering Geophysics, 5, no. 7, doi: 10.4133/JEEG5.1.7.

Sherrod, B. L., R. J. Blakely, C. S. Weaver, H. M. Kelsey, E. Barnett,L. M. Liberty, K. L. Meagher, and K. M. Pape, 2008, Findingconcealed active faults: extension of the southern Whidbey Islandfault across the Puget Lowland, Washington: Journal of Geophysi-cal Research, 113, B5, B05313, doi:10.1029/2007JB005060.

Wood, S. H., 1994, Seismic expression and geological signicance ofa lacustrine delta in Neogene deposits of the western Snake Riverplain, Idaho: AAPG Bulletin, 78, no. 1, 102121.

Xia, J. and R. D. Miller, 2000, Design of a hum lter for suppressingpower-line noise in seismic data: Journa l of Environmental and En-gineering Geophysics, 5, no. 2, 3138, doi:10.4133/JEEG5.2.31.

Corresponding author: [email protected]

Documents

Lee 000146