Ground-penetrating radar profiles over multiple steel tanks: Artifactremoval through effective data processing

Jorge Luís Porsani1 and William A. Sauck2

ABSTRACT

This paper describes some ground-penetrating radar�GPR� results to characterize steel tanks buried in the subsoilof an urban center. Tanks were installed in the first Braziliangeophysical test site located at the Institute of Astronomy,Geophysics, and Atmospheric Science �IAG� of the Univer-sity of São Paulo �USP�. This paper also presents an effec-tive procedure for removal of hyperbolic artifacts associatedwith GPR reflections between multiple steel tanks. One hun-dred sixty-five GPR profiles of 200 MHz, 400 MHz, and500 MHz �shielded bistatic antennae� were measured. Thework was done in two distinct places: on the geophysical testsite of the IAG/USP in the city of São Paulo, Brazil, and in anarea formerly occupied by a truck cleaning, refueling, andmaintenance facility, located in Oscoda, northeastern Michi-gan, USA. Predictive deconvolution and Kirchhoff migra-tion were first used to improve vertical and lateral resolution.To minimize the hyperbolic artifacts caused by reflectionsbetween tanks, high-pass filtering was applied to data. Then,Hilbert transforms emphasize the location of the steel tanksin an energy cross-section form. Further GPR processing,such as vertical and depth slices, were very important to im-prove visualization and localization of steel tanks and pipesin the subsurface.

INTRODUCTION

GPR is a nondestructive electromagnetic method that uses high-frequency �10 MHz–2.5 GHz� radio waves, and has several appli-cations for shallow subsurface investigations, such as mineral explo-ration, geological, geotechnical, hydrogeological, environmental,and archaeological studies. The theoretical basis for this method can

be found in Ward and Hohmann �1987�, Davis and Annan �1989�,Daniels �1996�, and Porsani �1999�, among others.

In particular, GPR is efficient for geologic characterization of thesubsurface as well as for mapping of nongeologic objects, for exam-ple, detection of steel, plastic, and concrete targets buried in the sub-soil �Daniels et al., 1995; Zeng and McMechan, 1997; Grandjean etal., 2000; Paniagua et al., 2004; Borges and Porsani, 2005; Lima andPorsani, 2005; Porsani et al., 2006; Rodrigues and Porsani, 2006�.

Rapid industrial development accompanying population growthin large cities causes various problems such as environmental con-tamination. Many other problems in urban areas are caused by sub-way line construction, the placing of pipes underground for watersupply, galleries for storm water drainage, high-voltage electric ca-bles, fiber-optic cables, and gas pipelines in the subsoil. The usualmain problems relate to the destruction of preexisting undergroundutilities, which disrupt the local economy, transport, communica-tion, and may even put lives at risk.

The use of geophysical methods, particularly GPR, before begin-ning geotechnical excavations in the great urban centers becomesimportant to detect previously installed utilities in the subsoil, suchas electric and telephone cables, water and gas pipelines, etc. There-fore, such methods can serve to prevent accidents and minimize therisks to society. However, data interpretation can be complex be-cause of ambiguities found in the results. To reduce the uncertain-ties, one line of research with the GPR method performs studies un-der controlled conditions aimed at better understanding the normalreflection patterns caused by different targets in the subsurface�Grandjean et al., 2000; Gerber et al., 2004; Paniagua et al., 2004;Porsani et al., 2006; Rodrigues and Porsani, 2006�.

In this context, this article presents some GPR 2D/3D results per-formed in a geophysical test site at the Institute of Astronomy, Geo-physics, and Atmospheric Science �IAG� at the University of SãoPaulo �USP�, where multiple steel tanks are buried, and in an areaformerly used for truck fueling and maintenance located in the townof Oscoda, northeastern Michigan, USA.

Manuscript received by the Editor 9August 2006; revised manuscript received 6 July 2007; published online 3 October 2007.1Universidade de São Paulo Instituto de Astronomia, Geofísica e Ciěncias Atmosféricas, Departamento de Geofísica, São Paulo, Brazil. E-mail: porsani@

iag.usp.br.2Western Michigan University, Department of Geosciences, Kalamazoo, Michigan. E-mail:sauck@wmich.edu.

© 2007 Society of Exploration Geophysicists.All rights reserved.

GEOPHYSICS, VOL. 72, NO. 6 �NOVEMBER-DECEMBER 2007�; P. J77–J83, 11 FIGS.10.1190/1.2783412

J77

As is already well known, GPR reflection from metal tanks ischaracterized by strong hyperbolic reflection patterns �Ulriksen,1982; Zeng and McMechan, 1997; among others�. However, elec-tromagnetic wave reflections between a pair of steel tanks cause athird hyperbola �an artifact�. This fact can be a pitfall in mappingsubsurface objects, because it causes an overestimation of targetnumbers in the subsurface. Therefore, to avoid an erroneous inter-pretation of the number and location of multiple targets in the sub-surface, this extra hyperbola �artifact� must be corrected or removed.

In this article, we show an effective procedure for removal of hy-perbolic artifacts caused by multiple steel tanks by using high-passfiltering where relevant position and depth information is retained.Predictive deconvolution and Kirchhoff migration were used to im-prove vertical and lateral resolution.

Knowledge of the correct spatial distribution of targets in the sub-soil is important in mapping buried utilities for urban planning andenvironmental engineering. This information serves as basis forplanning the advance of geotechnical excavations in urban centersbecause it helps minimize problems related to the damage or de-struction of preexisting underground utilities, and avoid dangerousaccidents.

STUDY AREAS

IAG/USP test site, São Paulo, Brazil

The test site at the Institute ofAstronomy, Geophysics, andAtmo-spheric Science �IAG� at the University of São Paulo �USP� is a pio-neering effort in Brazil, and it is located in a geologic-urban contextdifferent from other test sites set in Europe and the United States.This geophysical test site is situated at the border of São Paulo sedi-mentary basin, located in the city of São Paulo, Brazil �Figure 1�.

Aclayey-sand soil �typically tropical soil different from Europeanand North American soils� and sand-clayey sediments of the Re-sende and São Paulo formations underlie the local area, overlappingonto the granite-gneissic basement �Porsani et al., 2004�.

The test site was constructed for teaching and research. Construct-ed close to IAG/USP, it has an area of 1500 m2 �50 m�30 m�. Thetargets were grouped by types of materials, and buried along 7 linesin the magnetic north-south direction.Along metallic pipe of 3.8 cmdiameter was buried in an east-west orientation at 0.5-m depth at the15 m north position, crossing all lines, and serves as a guide or refer-ence target for various geophysical surveys.

Targets with different geometries were buried with long axes innorth-south or east-west directions, and at depths to tops varying

from 0.5 to 2 m. Details of its design and construction, as well assome GPR results, are found in Porsani et al. �2006�.

All materials were chosen to have magnetic and conductive prop-erties, or to generate reflections in GPR profiles in order to simulatethe main targets that usually are found in the urban center environ-ment and at some archaeological sites. After targets were buried, thelocal landscape was restored. The only evidence at the surface is fourconcrete marks placed at the corners of the test site.

As already mentioned, the IAG/USP test site is an important toolfor teaching and research in applied geophysics.At present, it is usedto check the penetration and resolution of different geophysicalmethods for detection and characterization of shallow targets�Borges and Porsani, 2005; Lima and Porsani, 2005; Porsani et al.,2006; Rodrigues and Porsani, 2006�.

In this article, some GPR results for characterizing multiple steeltanks are described. Cylindrical tanks of 200 liters were installed inhorizontal and vertical positions along line 4 at the IAG/USP testsite. The tanks, 0.59 m in diameter and 0.86 m long, were paintedwith antirust paint for longer life and buried empty to prevent corro-sion. The tanks were buried individually and in pairs, and they wereinstalled at 0.5-, 1.0-, and 2.0-m depths.

Figure 2 shows a photograph of the system for measuring the spa-tial positions of targets while they were being installed. Individualtanks and the steel reference pipe are spaced 4 m apart and the tankpair are spaced 1 m apart. The layout of steel tank positions alongline 4 will be presented in the analysis of results. This line was forsimulating an environmental site in which steel tank location is thegoal.

Oscoda, Michigan site, USA

Acommercial trucking company once owned this site where truckcleaning, refueling, and maintenance took place. It is located in the

Mato Grossode Sul

Paraná

Brazil

52°

44°0 100km

21°

23°

25°CenozoicParaná basinCrystalline basement

São PauloState

São Paulo

Atlantic

Ocean

MinasGerais

Figure 1. Location map of São Paulo city, São Paulo, Brazil.

0

1

Depth(m)

1 m

Figure 2. Installing and surveying locations of steel tanks inside theIAG/USP test site.

J78 Porsani and Sauck

town of Oscoda, northeastern Michigan, north of Saginaw Bay inIosco County. The site is on the sand-dominated coastal plain ofLake Huron, where sand is in excess of 15 m thick, and the water ta-ble lies at about 2-m depth.

The large steel underground storage tanks are located approxi-mately 100 m west of the shoreline of Lake Huron, one of the GreatLakes. The location is at present the unpaved sand-surfaced portionof a parking lot behind a bowling alley, near the property line with anadjacent restaurant’s parking area. No structures remain from theformer use of this site.

ACQUISITION AND PROCESSING OF GPR DATA

One hundred sixty-four GPR profiles of 200 MHz and 400 MHzwere acquired at the IAG/USP geophysical test site, São Paulo, Bra-zil, and one profile of 500 MHz was acquired at the Oscoda site. Inall profiles, shielded bistatic antennae and GSSI Model SIR-3000�USP� and SIR-10A� �Oscoda� equipment were used. At the testsite, the profiles were spaced 1 m apart, the interval between traces is5 cm, and they were acquired along both north-south and east-westdirections. Although this is acceptable for 3D imaging of 2D objects�e.g., a pipe�, aliasing will occur for 3D objects smaller than a wave-length.

To avoid aliasing and to obtain 3D full-resolution GPR imaging,the interval between traces needs to be reduced �Grasmueck et al.,2005�. In the Oscoda area, the profile was acquired using an intervalbetween the traces of 1.5 cm.At both sites, profiles were acquired byusing continuous mode, stack of 4, and electric field-Ey polarization�long antenna axis perpendicular to transect direction�.

GPR processing was done using Radan software �GSSI�. The ob-jectives of GPR data post-processing were to improve vertical reso-lution and spatial positioning of the tanks. The main stages used instandard processing were: time zero-correction, time filtering �band-pass�, and time-varying gains. Band-pass filtering was applied to re-duce the amplitude of low and high frequency noise. Time-varyinggains were applied to compensate for absorption, spherical diver-gence, and signal scattering. This standard procedure does not re-move the hyperbolic artifact that is located below and midway be-tween the hyperbolas from the pair of tanks.

In order to explain the origin of the artifact, Figure 3a shows viapositions A, B, and C a schematic GPR raypath over two steel tanksburied at the IAG/USP test site. Figure 3b shows the simulated re-sults of wave propagation considering zero-offset between transmit-ter and receiver antennas, normal incidence, 0.97-m depth for tank-1�h1�, 0.99-m depth for tank-2 �h2�, 1.0 m distance �d� between tanks,0.3-m ray �r�, 0.06-m/ns velocity �v�, and position �x� in the distanceaxis. Simplified A, B, and C equations for the raypath are presentedat the top of Figure 3b.

To remove hyperbolic artifacts caused by reflections between thetanks, additional processing was applied to the data. The main stagesused in this effective processing were: predictive deconvolution,high-pass filtering varying in time, and Kirchhoff migration and Hil-bert transforms. In both standard and effective processing, time/depth conversion was done. Vertical axes show depth �left� and two-way traveltime �right�.

Predictive deconvolution �for instance, operator length � 21,prediction lag � 12, prewhitening � 10, and samples � 35–512�was used to remove periodic or reverberatory GPR reflection eventsfrom the steel tanks, and to improve the vertical resolution throughpulse compression �Robinson and Treitel, 1980; Yilmaz, 1987�.

High-pass filtering was applied in the entire time window of GPRprofiles to improve the signal-to-noise ratio and reduce the ampli-tude of the hyperbolic artifact. In order to remove or minimize the in-fluence of artifacts, high-pass filtering was applied again, in whichthe start of the time window corresponds to the top of the artifact�40 ns� and extends to the end of the trace.

The lateral positioning of multiple steel tanks was improved usingKirchhoff migration to collapse the hyperbolae. For GPR data mi-gration from line 4 of the IAG/USP test site, a variable apparent ve-locity model, such as v1 � 0.060 m/ns for the 0–20 ns interval, andv2 � 0.067 m/ns for times �20 ns, was used.

For migration of the profile from Oscoda, Michigan, a constantapparent velocity of 0.155 m/ns was applied. At the IAG/USP testsite, as well as the Oscoda site, these velocities were obtained by an-alyzing the geometry of the hyperbolae associated with the reflectionof various steel tanks, and that of the steel reference pipe �IAG/USP�, being calculated by the x2 � t2 method �Dix, 1955�.

Hilbert transforms were used to calculate the amplitude envelopeof traces along the GPR profiles, to better emphasize the location ofthe steel tanks and the reference pipe in energy cross-section form.Finally, for conversion from time to depth in the test site profiles, adielectric constant model of 25 for the 0–20-ns interval, and 20 fortimes �20 ns, was used to be compatible with wet clayey-sand soil�Porsani et al., 2004�.

In contrast, in profiles at Oscoda, a dielectric constant of 3.7 wasused to be compatible with dry sandy soil. GPR results, in the form ofvertical and depth slices, were very important to improve visualiza-tion and characterization of steel tanks, and the steel reference pipein the subsurface.

ANALYSIS OF RESULTS

IAG/USP test site, São Paulo, Brazil

Figure 4 shows a 200-MHz GPR profile over paired steel tanks�displayed in Figure 3a� buried at the IAG/USP test site. Figure 4awas obtained after standard data processing without predictive de-convolution. In this figure, we can see the margins of the excavatedtrench to install the tanks, and the tops of the two steel tanks �1 and 2�

a)0.0

0.5

1.0

1.5

2.0

Position (m)

Depth(m)

0 1 2 3 4 5

21B

Tank Tank

A C

b)0.0

0.5

1.0

1.5

2.0

0

10

20

30

40

50

60

70

Position (m)

Depth(m)

Two-waytraveltime(ns)

0 1 2 3 4 5

A

A + B + Cartifact

A = 2(√x2 + (h1 + r)2 – r)/v

C

C = 2(√x— d)2 (h2 + r)2 – r)/v

B = (A + C)/2 + (d – 0,6r)/v

Figure 3. �a� Schematic GPR raypath between two buried steel tanks.�b� Simulation of wave propagation, considering zero offset be-tween transmitter and receiver antennas, where the simplified equa-tions for the A, B, and C raypaths are presented at the top of thefigure.

Effective processing of GPR data J79

characterized by strong hyperbolic reflectors. Note that a third hy-perbolic reflector appears lower, between the tanks. This is the arti-fact caused by wave reflections from tank 1 to tank 2 before it returnsto the surface �raypath B in Figure 3a and A � B � C hyperbolasimulation results in Figure 3b�.

Figure 4b shows the GPR profile after Kirchhoff migration fromFigure 4a. In this figure, hyperbolic reflectors for tank 1 and tank 2were collapsed to a point, and the artifact changed to a “smile” arti-fact. The smile artifact appears because the apparent velocity mea-sured for hyperbolae of tanks 1 and 2 �0.067 m/ns� is 15% greaterthan the apparent velocity measured for the hyperbola artifact�0.057 m/ns� �Yilmaz, 1987�. Therefore, to improve GPR profilequality over adjacent steel tanks, the hyperbolic artifacts must becorrected or removed.

Figure 5 presents a 200-MHz GPR profile over paired steel tanksburied at the IAG/USP test site after effective processing data fromFigure 4a. Results obtained after deconvolution, and high-pass fil-tering appear in Figure 5a. We can observe that the hyperbolic arti-fact �Figure 4a� was removed, whereas hyperbolic reflections be-cause of the tanks �1 and 2� became clearer, and reverberations of the

electromagnetic signal became attenuated. This is because the high-pass filter removes the lower-frequency artifact, and because the de-convolution filter increases the vertical resolution �Yilmaz, 1987�.

Figure 5b shows the GPR profile after Kirchhoff migration fromFigure 5a. Observe that the hyperbolic reflectors for tanks 1 and 2were collapsed to a point and the smile artifact �see Figure 4b� disap-peared completely. Therefore, these results showed that the use ofeffective processing was very important to improve the quality of theimage, increase the vertical resolution, and permit the almost com-plete removal of this artifact.

To improve the resolution of hyperbolic reflectors over two steeltanks, and to show again the hyperbolic artifact as a result of the re-flection between the tanks, a 400-MHz GPR profile was acquiredover the same tanks displayed in Figure 3a. The two stages of pro-cessing, standard and effective, also were done.

Figure 6 shows the 400-MHz GPR profile over two steel tanksburied at the IAG/USP test site. Figure 6a was the result after thestandard processing data without predictive deconvolution. Stronghyperbolic reflectors characterize the tops of the two steel tanks �1and 2�. The third hyperbolic reflector �marked with an arrow in Fig-ure 6a� is clearer than the one observed in Figure 4a, and it is the arti-fact caused by the wave reflections between tank 1 and tank 2 ofFigure 3.

Figure 6b shows the GPR profile after migration of Figure 6a.Note that the hyperbolic reflectors for tanks 1 and 2 were nearly col-lapsed to a point and that the hyperbolic artifact changed to a smileartifact in Figure 4b.

Figure 7 shows the 400-MHz GPR profile over the paired steeltanks buried at the IAG/USP test site after additional �effective� pro-cessing of data from Figure 6a. Results obtained after deconvolutionand high-pass filtering are shown in Figure 7a. Observe that the hy-perbolic artifact �Figure 6a� was removed, the hyperbolic reflectionsappearing at the tops of tanks were clearer, and reverberations ofelectromagnetic signal were attenuated. Figure 7b shows the GPRprofile after migration from Figure 7a. Hyperbolic reflectors for thetwo tanks were collapsed to a point, and the smile artifact �seen inFigure 6b� was completely removed.

Results obtained for 200-MHz and 400-MHz GPR profiles overthe two steel tanks were encouraging. Subsequently, effective pro-cessing stages were applied to line 4 of the IAG/USPtest site in orderto improve the quality of the entire GPR profile, and to best charac-terize all of the steel tanks.

2

b)0.0

0.5

1.0

1.5

2.0

0

10

20

30

40

50

60

70

Position (m)

Two-waytraveltime(ns)

0 1 2 3 4 5a)0.0

0.5

1.0

1.5

2.0

0

10

20

30

40

50

60

70

Position (m)

Depth(m)

0 1 2 3 4 5

Smileartifact

Artifact

Trench

1

Figure 4. The 200-MHz GPR profile over a pair of steel tanks buriedat the IAG/USPtest site. �a� Standard processing without deconvolu-tion. �b�After Kirchhoff migration from �a�.

b)0.0

0.5

1.0

1.5

2.0

0

10

20

30

40

50

60

70

Position (m)

Two-waytraveltime(ns)

0 1 2 3 4 5a)0.0

0.5

1.0

1.5

2.0

0

10

20

30

40

50

60

70

Position (m)

Depth(m)

0 1 2 3 4 5

Figure 5. 200-MHz GPR profile over a pair of steel tanks buried atthe IAG/USP test site. �a� Effective processing after deconvolutionand high-pass filtering. �b�After Kirchhoff migration from �a�.

b)0.0

0.5

1.0

1.5

0

10

20

30

40

50

60

Position (m)Two-waytraveltime(ns)

0 1 2 3 4 5a)0.0

0.5

1.0

1.5

0

10

20

30

40

50

60

Position (m)

Depth(m)

0 1 2 3 4 5

SmileartifactArtifact

Figure 6. 400-MHz GPR profile over paired steel tanks buried at theIAG/USP test site. �a� Standard processing without deconvolution.�b�After Kirchhoff migration from �a�.

J80 Porsani and Sauck

Figure 8 shows tank locations and 200-MHz GPR results overmultiple steel tanks buried at various depths in line 4 from the testsite. Figure 8a graphs a section to show the spatial distribution of200-liters steel tanks. Figure 8b shows the GPR profile after decon-volution and high-pass filtering �blue rectangle shows an enlargedarea and with depth slices to be shown in Figure 9�. Strong hyperbol-ic reflectors characterize tank tops as a consequence of the high-con-ductivity steel �Olhoeft, 1998�.

Figure 8c shows the migrated profile. Observe that the hyperbolaswere collapsed to a point and the tanks are focused. Figure 8d showsthe profile after applying the Hilbert transform. Beyond a 2.5-mdepth, the GPR signal is strongly attenuated because of the high con-ductivity of sediments of the São Paulo basin �Porsani et al., 2004�.

Figure 9 details part of Figure 8b. The figure shows a 3D diagramwith a horizontal slice at 0.7-m depth and a vertical slice along thetank centers. These results showing horizontal and vertical slices areimportant because they improve visualization and characterizationof the steel tanks and steel reference pipe. In this figure, the GPR sig-nature is clear for steel targets in terms of reflection and signal ampli-tude. Both targetAand the linear anomaly in the slice at 0.7-m depthcorrespond to the steel reference pipe that crosses all lines of the testsite. Hyperbolic B, C, and D reflections, all corresponding to steeltanks, are shown in Figure 8b.

Results obtained with effective processing stages for 200-MHzand 400-MHz GPR profiles over multiple steel tanks buried at theIAG/USP test site were very interesting. The results show improveddetection of steel tanks, and artifact removal. In order to verify theeffectiveness of the processing, additional processing stages wereapplied to another GPR profile acquired in a very different place, asdescribed next.

Oscoda site, Michigan, USA

Figure 10 shows a 500-MHz GPR profile over multiple steel tanksburied in Oscoda, Michigan, USA. Figure 10a was obtained after thestandard processing without predictive deconvolution. Strong hy-perbolic reflectors characterize the tops of the two steel tanks. Notethat between the 1- and 2-m positions at a depth less than 0.5 m, twolittle hyperbolic reflectors appear. These reflectors are caused by twofiller or vent pipes over the first tank. Note that these tanks are morethan two orders of magnitude larger in volume than the barrels at theIAG/USP site discussed earlier.

b)0.0

0.5

1.0

1.5

0

10

20

30

40

50

60

Position (m)

Two-waytraveltime(ns)

0 1 2 3 4 5a)0.0

0.5

1.0

1.5

0

10

20

30

40

50

60

Position (m)

Depth(m)

0 1 2 3 4 5

Figure 7. 400-MHz GPR profile over paired steel tanks buried at theIAG/USP test site. �a� Effective processing after deconvolution andhigh-pass filtering. �b�After Kirchhoff migration from �a�.

a)

b)

01234

Distance (m)

Legend: Steel tank horizontalSteel tank vertical

Steel reference pipe

Depth(m) 0

1234 D

epth(m)

0 10 20 30

0

1

2Depth(m) 0

20406080 Ti

me(ns)

0 10 20

BA

C D

30

c)0

1

2Depth(m) 0

20406080 Ti

me(ns)

0 10 20 30

d)0

1

2Depth(m) 0

20406080 Ti

me(ns)

0 10 20 30

S N

Figure 8. 200-MHz GPR profiles over all of the steel tanks buriedin line 4 of the IAG/USP test site. �a� Section showing spatial distri-bution of 200-liter steel tanks. �b� Profile after deconvolution andhigh-pass filtering �blue rectangle shows enlarged area and withdepth slices in Figure 9. �c� Migrated profile. �d� Profile after Hilberttransforms.

0.0

0.5

1.0

1.5

Distance (m)Distance (m)

A

B

C D

Depth(m)

3028

2624

1015

20

Figure 9. Detailed area of Figure 8b �200 MHz after deconvolutionand high-pass filtering�, showing a vertical section and depth slice at0.7 m over tanks in the center of line 4.

b)0.00.51.01.52.02.53.0

0

10

20

30

40

Position (m)

Two-waytraveltime(ns)

a)0.00.51.01.52.02.53.0

0

10

20

30

40

Position (m)

Depth(m)

0 1 2 3 4 5 6 0 1 2 3 4 5 6

SmileartifactArtifact

Figure 10. 500-MHz GPR profile over two large steel tanks buried inOscoda, northeastern Michigan, USA. �a� Standard processing with-out deconvolution. �b�After Kirchhoff migration from �a�.

Effective processing of GPR data J81

The third hyperbolic reflector �marked with an arrow in Figure10a� is the artifact caused by the wave reflections between thetanks. Figure 10b shows the GPR profile after Kirchhoff migration�0.155 m/ns� of Figure 10a. Note that the hyperbolic reflectors forthe two tanks were collapsed to a point, and the hyperbolic artifactchanged to a smile artifact.

Figure 11 shows the same 500-MHz GPR profile after applyingeffective processing to Figure 10a. Results obtained after deconvo-lution and high-pass filtering are shown in Figure 11a. Note that thehyperbolic reflections from the tank tops were clearer, electromag-netic signal reverberations were attenuated, and the hyperbolic arti-fact �see Figure 10a� was minimized. Figure 11b shows the GPR pro-file after migration of Figure 11a. Hyperbolic reflectors for the twotanks were collapsed to two points, and the smile artifact �see Figure10b� had been removed.

CONCLUSIONS

The IAG/USP geophysical test site constructed on the Universityof São Paulo campus is a pioneering effort in Brazil. Various steel,plastic, and concrete targets were buried in order to simulate geo-technical problems, objects found during environmental studies, andobjects typical of archaeological sites. At first, the test site was con-structed just for teaching and research purpose.At present, the site isbeing used to check the penetration and resolution of different geo-physical methods to detect and characterize shallow targets.

The effective processing of 200-MHz, 400-MHz, and 500-MHzGPR profiles over multiple steel tanks recorded over the IAG/USPtest site �São Paulo, Brazil� and in an abandoned area of truck fuelingand maintenance �Michigan, USA� improved the quality of signaland vertical resolution, enhanced lateral tank positioning, and re-moved hyperbolic artifacts caused by reflections between tanks.Ad-ditional 3D processing, such as vertical and depth slices, improve vi-sualization and characterization of the steel tanks, and of a steel ref-erence pipe buried in the subsurface.

Geophysical signatures of known targets buried at the IAG/USPtest site contribute to our ability to interpret geophysical data by theiruse as typical examples, and could be extrapolated to areas where in-formation about the subsurface was lacking. We found that the sameeffective processing or enhancement techniques developed for thesmall drums at the IAG/USPtest site were applicable to far larger un-derground storage tanks at the field site in Oscoda, Michigan.

ACKNOWLEDGMENTS

The first author is grateful to Fundação de Amparo à Pesquisa doEstado de São Paulo �FAPESP�, a Brazilian state-level researchagency, for providing financial support for construction of the IAG/USP test site �Grant 02/07509-1�, and for his post-doctoral fellow-ship at Western Michigan University in 2005 �Grant 04/13609-4�.IAG/USP is acknowledged for providing infrastructure support. Theauthors thank colleagues from the Geophysics Department forproject contributions, as well as for assistance in test site construc-tion. We thank Ernande C. Santos and students Welitom R. Borges,Selma I. Rodrigues, Cláudia de Oliveira, Robson S. Lima, ViníciusR. N. dos Santos, Elizete M. Silva, and Marcus V. Lima, among oth-ers for helping in the installation, and for geophysical data acquisi-tion before and after the targets were installed. Finally, we thank thereviewers for their valuable critiques and suggestions which greatlyimproved the paper.

REFERENCES

Borges, W. R., and J. L. Porsani, 2005, Utilização de metodologias geofísicasna identificação de manilhas de concreto instaladas no sítio controlado degeofísica rasa do IAG/USP: Proceedings of the 9th International Geophys-ics Congress of the Brazilian Geophysical Society �SBGf�, Expanded Ab-stract, CD-ROM.

Daniels, J. J., R. Roberts, and M. Vendl, 1995, Ground penetrating radar forthe detection of liquid contaminants: Journal of Applied Geophysics, 33,95–207.

Daniels, D. J., 1996, Surface penetrating radar: IEE.Davis, J. L., and A. P. Annan, 1989, Ground penetrating radar for high resolu-

tion mapping of soil and rock stratigraphy: Geophysical Prospecting, 37,531–551.

Dix, C. H., 1955, Seismic velocities from surface measurements: Geophys-ics, 20, 68–86.

Gerber, R., P. F. Henningsen, C. Salat, and A. Junge, 2004, Investigation ofthe GPR reflection pattern for shallow depths on a test site: Proceedings ofthe 10th International Conference on Ground Penetrating Radar, Expand-edAbstracts, 275–278.

Grandjean, G., J. C. Gourry, and A. Bitri, 2000, Evaluation of GPR tech-niques for civil-engineering applications: Study on a test site: Journal ofApplied Geophysics, 45, 141–156.

Grasmueck, M., R. Weger, and H. Horstmeyer, 2005, Full-resolution 3DGPR imaging: Geophysics, 70, 12–19.

Lima, R. S., and J. L. Porsani, 2005, Localização de manilhas de concreto in-staladas no sítio controlado de geofísica rasa do IAG-USP através demulti-configurações das antenas GPR: Resultados preliminares: Proceed-ings of the 9th International Congress of the Brazilian Geophysical Soci-ety �SBGf�, ExpandedAbstracts, CD-ROM.

Olhoeft, G., 1998, Electrical, magnetic, and geometric properties that deter-mine ground penetrating radar performance: Proceedings of the 7th Inter-national Conference on Ground Penetrating Radar, Expanded Abstracts,177–182.

Paniagua, J., M. Del Rio, and M. Rufo, 2004, Test site for the analysis of sub-soil GPR signal propagation: Proceedings of 10th International Confer-ence on Ground Penetrating Radar, ExpandedAbstracts, 751–754.

Porsani, J. L., 1999, Ground penetrating radar �GPR�: Proposta metod-ológica de emprego em estudos geológico-geotécnicos nas regiões de RioClaro e Descalvado—SP: D.Sc. thesis, Universidade Estadual PaulistaJúlio de Mesquita, Brazil.

Porsani, J. L., W. R. Borges, V. R. Elis, L. A. Diogo, F. Y. Hiodo, A. Marrano,and C. A. Birelli, 2004, Investigações geofísicas de superfície e de poço nosítio controlado de geofísica rasa do IAG-USP: Revista Brasileira deGeofísica, 22, 3, 245–258.

Porsani, J. L., W. R. Borges, S. I. Rodrigues, and F. Y. Hiodo, 2006, O sítiocontrolado de geofísica rasa do IAG-USP: Instalação e resultados GPR2D-3D: Revista Brasileira de Geofísica, 24, 1, 49–61.

Robinson, E. A., and S. Treitel, 1980, Geophysical signal analysis: Prentice-Hall.

Rodrigues, S. I., and J. L. Porsani, 2006, Utilização do GPR para caracterizartambores plásticos enterrados no sítio controlado de geofísica rasa doIAG/USP: Revista Brasileira de Geofísica, 24, 2, 157–168.

b)0.00.51.01.52.02.53.0

0

10

20

30

40

Position (m)

Two-waytraveltime(ns)

a)0.00.51.01.52.02.53.0

0

10

20

30

40

Position (m)

Depth(m)

0 1 2 3 4 5 6 0 1 2 3 4 5 6

Figure 11. 500-MHz GPR profile over steel tanks buried in Oscoda,northeastern Michigan, USA. �a� Effective processing after decon-volution and high-pass filtering. �b� After Kirchhoff migration from�a�.

J82 Porsani and Sauck

Ulriksen, P., 1982, Application of impulse radar to civil engineering: Ph.D.thesis, University of Technology, Lund.

Ward, S. H., and G. W. Hohmann, 1987, Electromagnetic theory for geophys-ical applications, in M. N. Nabighian, ed., Electromagnetic methods in ap-

plied geophysics: SEG, 1, 131–311.Yilmaz, O., 1987, Seismic data processing: SEG.Zeng, X., and G. A. McMechan, 1997, GPR characterization of buried tanks

and pipes: Geophysics, 62, 797–806.

Effective processing of GPR data J83