Site characterization at the GroundwaterRemediation Field Laboratory

WILLIAM P. CLEMENT, STEVE CARDIMONA, ANTHONY L. ENDRES, Boston College, Boston, Massachusetts KATHARINE KADINSKY-CADE, Phillips Laboratory, Hanscom Air Force Base, Massachusetts

nvironmental site assessment usu-ally includes determining the geo-logic character of the subsurface aftercontamination has occurred. How-ever, at the Groundwater Remedia-tion Field Laboratory (GRFL), weacquired extensive geophysical andgeotechnical data before planned con-taminant releases happened. Theseprespill geophysical images providebackground data for comparisonwith data acquired in the future.Additionally, a large amount of conepenetrometer (CPT) data provide insitu geotechnical measurements tocompare with surface geophysics.The end product is a unique data setand an extensive analysis, resultingin a detailed description of the sub-surface based on standard geophys-ical and geotechnical methods.

Site characterization for environ-mental cleanup provides the infor-mation needed to determine theextent and scope of a specific prob-lem and to design remediationstrategies. This can be expensive andhazardous when based primarily onin situ sampling by invasive tech-niques, such as monitoring wells orgeologic coring. For example, bore-holes must be closely spaced toinsure that important features arefound. Furthermore, drill holes mayremobilize contaminants and pro-vide conduits that may lead to fur-ther contamination. And, monitor-ing wells must be sampled for yearsto comply with environmental regu-lations. To reduce costs and to pro-vide characterization over a widerarea, geophysical methods often sup-plement drilling. Unfortunately,interpretation of surface geophysicaldata in terms of relevant physicalproperties is often difficult becauseof the scarcity of direct subsurfacecontrols. Such controls, though, arepresent at the GRFL and provide anopportunity to study the relationshipbetween in situ geotechnical mea-surements of physical propertieswith a variety of geophysical data.

The GRFL, administered by theU. S. Air Force’s Armstrong Labora-

tory under the Strategic Environ-mental Research and DevelopmentProgram, is located at Dover AirForce Base, Dover, Delaware.

Various groups conducted com-plementary geophysical surveys ofthe site during 1995 and 1996, col-lecting ground penetrating radar(GPR) with pulseEKKO and SIRunits at multiple source frequencies,shallow high resolution reflectionand refraction seismic data, one-

dimensional electrical resistivity sur-veys, and terrain conductivity mea-surements (Figure 1). We recordedhigh-resolution seismic reflectionlines along north-south profiles sep-arated by 10 m. Lines in a GPR gridwere also separated by 10 m. Thenorth-south seismic reflection lineswere coincident with some GPRlines. Applied Research Associatessimultaneously collected soil sam-ples and performed CPT “pushes.”

NOVEMBER 1997 THE LEADING EDGE 1617

E

Figure 1. Location of some geophysical surveys. Hexagons = locations ofthe electrical resistivity soundings; colors are coordinated with Figure 6.Gray dots = locations of CPT pushes. Dashed line = the seismic reflectionprofile in Figure 2. The 20 MHz GPR profile in Figure 3 is coincident withthis line. Dotted line = the 50 MHz GPR profile of Figure 4. Solid line =the seismic refraction survey in Figure 5. Terrain conductivity measure-ments cover the field (see Figure 7).

Several shallow wells, previous-ly drilled near the site, provide adetailed look at the medium tocoarse grained sands and silt thatcomprise the local stratigraphy. Anunconfined aquifer (water table at7.3 m and bounded below by a clayaquitard whose depth is 9-14 m)defines the main hydrologic feature.Three main features of the aquiferprovide targets for the geophysicalsurveys. The primary target is thesand-clay interface that marks theaquitard boundary at the base of thesurface aquifer; the second is thewater table; the third is a thin, shal-low clay layer about 4 m below thesurface in the northern portion ofthe GRFL. Imaging this clay layer,

which was discovered by the CPTsurveys, tests our ability to geo-physically determine aquifer het-erogeneity. The different geophysi-cal techniques image these featureswith varying degrees of success.

Seismic and GPR. GPR works bestin electrically resistive environmentssuch as dry, unconsolidated sandswith little clay. High resolution seis-mic reflection works best wherehigh frequencies from the sourcepropagate efficiently in the earth,such as in water-saturated condi-tions or in clay-rich materials. Notsurprisingly, when one of these tech-niques provides high quality data,the other usually does not.

At the GRFL, we obtained gooddata from both techniques and weidentified common reflections onthe respective sections. A reflectionis observed at 40-50 ms across mostof the seismic reflection profile (Fig-ure 2). In the GPR image (Figure 3),a reflection is observed at 200-300ns. Using velocities determinedfrom the GPR data, the reflector isabout 11 m below the surface in thesouth and deepens to about 15 m inthe middle of the section. We inter-pret this as the reflection from thebase of the aquifer. The depth to thereflector in the seismic section is thesame. These depths agree with thedepth to the aquitard measured inthe CPT surveys. Clearly, GPR andhigh resolution seismic adequatelyimage the base of the aquifer.

In the seismic data, a shallowerreflection at 20-30 ms is easily seenat 90-165 m (along the distance axis).Using the stacking velocities to esti-mate depth in the time section, thisreflector is about 4 m deep and coin-cides with the northern shallow claylayer observed in the CPT data. Areflection from the shallow clay isnot observed in the 20 MHz GPRdata. However, in the 50 MHz GPRdata (Figure 4), a reflection exists atabout 4 m, because the shorterwavelengths in the higher frequen-cy data can better resolve thinnerlayers. Again, the seismic and GPRdata appear to image the same sub-surface feature.

The first events in the GPRimages are the air and groundwaves. Unlike seismic velocities, EMvelocities usually decrease withdepth and the EM velocity in air, thespeed of light, is much faster than inother materials. Thus, the wavestraveling directly between theantennas are the first to arrive. Also,diffraction hyperbolas from nearbylight towers and fences are easilyseen in the GPR images. These con-ductive, metallic objects are strongreflectors and EM waves propagatewith little attenuation in air, soreflections from above ground fea-tures are often present in GPR data.Even though GPR images appearsimilar to seismic reflection images,these differences are reminders thatthe two methods image differentphysical properties.

Seismic refraction is the onlytechnique to image the water table(Figure 5). All other methods,including the in situ CPT probes,show little, if any, indication of it.

1618 THE LEADING EDGE NOVEMBER 1997

Figure 2. Seismic reflection data from the north-south profile along 40 mwest. The reflector between 40 and 50 ms corresponds to the aquitardboundary. The reflector between 20 and 30 ms on the northern half of thesection corresponds to a shallow clay layer.

Figure 3. 20 MHz SIR-2 data along the same profile as the seismic reflec-tion data in Figure 2. The prominent reflection between 200 and 300 nsthat ends at about 90 m is the aquitard reflection.

Seismic refraction is a reliablemethod to determine depths to largevelocity changes, such as depth-to-basement in aquifer studies. In the

seismic refraction data, the directwave has a slow velocity (about 400m/s) which is not unusual for dry,unconsolidated sands at or near the

surface. A refracted wave overtakesthe direct wave about 20 m from theshot point. Based on a simple one-dimensional interpretation, the inter-cept time indicates that the refractionis from a depth of about 8 m — theapproximate depth of the watertable. Additionally, the refraction hasan apparent velocity of 1600 m/s.This indicates that the refraction isfrom the top of the water table.

Electrical methods. We used twomethods to measure electrical prop-erties at the GRFL. The electricalmethods complement GPR sincethey measure the resistivity or con-ductivity of the subsurface, whereasGPR data are primarily sensitive tochanges in dielectric permittivitybelow ground level.

Kick Geoexploration acquired 14 resistivity soundings using aSchlumberger array at locationsshown in Figure 1. Figure 6 showsapparent resistivity measurementsfrom six soundings conducted ap-proximately along a north-south pro-file at 60 m west. The sounding at thesouthernmost station has the highestapparent resistivity at AB/2 spacingsless than 50 m. As the survey pro-

NOVEMBER 1997 THE LEADING EDGE 1619

Figure 4. 50 MHz pulseEKKO IV data overlain by CPT soil classificationinterpretations. The color scale indicates the interpretation of the soil type.Data are from the north-south profile along 30 m west, 10 m to the east ofthe seismic reflection data in Figure 2.

Figure 5. Seismic refraction data along an east-west profile. The thin, solid lines indicate the direct wave and therefracted wave. The one-dimensional velocity profile is presented on the right and clearly shows the water table atabout 8 m. The velocity jump at 30 m is based on a weak, secondary arrival (not marked).

gresses to the north, apparent resis-tivities decrease, indicating that thesubsurface is becoming more con-ductive.

Terrain conductivity surveyswere acquired using EM-31 and EM-34-3 systems (Figure 7). They pro-vided depth penetration of 6-15 m.These indicate that the northern sec-tion is more conductive than thesouthern end, which agrees with theresistivity data. Although manythings can increase conductivity inthe earth, the electrical methods sug-gest the shallow, electrically conduc-tive clay layer first identified by CPTsampling.

Cone penetrometer. Applied Re-search Associates directly measuredsubsurface physical properties usingCPT surveys, in which an instrumentis pushed into the ground whilereadings are taken. CPT data are pre-sented in terms of measured or cal-culated properties as a function ofdepth (Figure 8). The piezoelectrictool is the basic instrument. Thisdevice measures the sleeve and tipstresses and the pore pressure. Withthese geotechnical quantities, we canuse empirical formulae to infer thesoil type.

Other instrument packages,including a resistivity tool and adielectric tool, were attached to theCPT stem. The former measures theresistivity with a small, four-elec-trode array. The dielectric probemeasures the permittivity using res-onant frequency. These data aredirect samples of the ground beneaththe GRFL and verify the surface geo-physics.

Figure 4 shows 50 MHz GPRdata overlain with soil classificationbased on interpretations of CPT data.The GPR reflection at 10-12 m corre-sponds with the top of the aquitard,a sand-to-clay transition. In theupper to middle section, the manyradar reflections show a complicatedsubsurface stratigraphy with thestructural trend dipping to the north.However, based on the CPT data, thestructural trend could be interpretedas dipping to the south. Also, manyreflections in the GPR data do notcorrespond to soil type changesinterpreted from the CPT data. Inter-ference of radar waves may causepart of the discrepancy. The CPTdata has a vertical sampling intervalof 5 cm, so some of the detail fromthe CPT is below the resolution levelof 50 MHz GPR. In general, the geo-

1620 THE LEADING EDGE NOVEMBER 1997

Figure 6. Resistivity soundings along a south-north profile near the west-ern edge of the GRFL showing decreasing resistivity to the north. The dataare plotted as apparent resistivities. AB is the separation distance of theelectrodes. The colors are coordinated with the electrical resistivity loca-tions in Figure 1. The increasing size of the symbols indicates increasingMN (potential electrode) separations.

Figure 7. Quadrature component of EM-31 (left) and EM-34 (right) terrainconductivity measurements. The EM-34 data are for a coil separation of 20m. The small dots on each plot are survey locations. The data contours arein mS/m. Each plot is color-scaled independently, but blues indicate rela-tively low conductivity and reds indicate relatively high conductivity,showing the higher conductivity in the north. The white dashed line is theknown extent of the shallow clay layer from CPT sampling.

physical and CPT interpretationsmatch regarding large-scale features,but much work remains to correlatethe physical properties measured orinferred by the different methods.

Conclusions. Our data from theGRFL show that the geophysical andgeotechnical methods image similarfeatures. Additionally, higher reso-lution methods like GPR show thecomplex, heterogeneous nature ofthe subsurface that interpretationsbased solely on a few wells or evenquite a few CPT pushes may over-simplify. Interestingly, previousgeological studies indicated a simpleand homogeneous subsurface be-neath Dover Air Force Base, a crite-ria for locating the GRFL. But subse-quently acquired geophysical data

clearly show the complicated natureof the shallow subsurface and thenecessity of more continuous sam-pling than is provided by in situmeasurements alone.

We continue to refine subsurfacemodels based on joint interpretationsof the GRFL data sets outlined in thispaper and data not presented, suchas high frequency GPR and datafrom other CPT tools. The comple-mentary nature of the geophysicalmeasurements should provide amore complete description of thesubsurface. We are continuing toinvestigate the relationship betweensurface geophysical measurementsand the in situ CPT measurements todetermine if methods measuringsimilar physical properties (e.g., elec-trical resistivity soundings and

CPT resistivity probes) result in sim-ilar subsurface models. Discrepan-cies in the details of separate inter-pretations may be due to unequalvertical and lateral averaging of the different techniques. We need abetter understanding of fundamentaldifferences such as this subsurfacevolumetric sampling before jointinterpretation can become jointinversion for geologic structure ofthe shallow subsurface.

Acknowledgments: We thank Rex Morey,Mark Noll, and the CPT crew of AppliedResearch Associates; John Madsen of the University of Delaware, Doria Kutrubes of Hager Geoscience; Jack Kick of Kick Geo-exploration; Richard Miller of the KansasGeological Survey; and Gary Olhoeft of the Colorado School of Mines for their helpacquiring the data. This work was sponsored(in part) by the Air Force Office of ScientificResearch, USAF, under LRIR #94PL001. The views and conclusions are those of the authors and should not be interpreted as necessarily representing the official policiesor endorsements, either expressed or implied,of the Air Force Office of Scientific Researchor the U.S. government. Much of the GPRdata was processed with Seismic Un*x.

Corresponding author: Bill Clement, 617-377-2611.

NOVEMBER 1997 THE LEADING EDGE 1621

Figure 8. CPT measurements of different physical properties. The fourupper plots are from a location within the shallow clay layer in the north-ern portion of the GRFL. The lower four plots are from the southern endof the GRFL away from the shallow clay layer. Note the presence of theshallow clay layer at a depth of 4-5 m in the upper plots. Ratio = correctedratio between sleeve stress and tip pressure; Class. = soil classificationindex; log (Resist.) = log (resistivity); and Dielect. = dielectric permittivity.

LE