lable at ScienceDirect

Journal of Archaeological Science 46 (2014) 96e106

Contents lists avai

Journal of Archaeological Science

journal homepage: http: / /www.elsevier .com/locate/ jas

Ground-penetrating radar investigations at Marj Rabba, a Chalcolithicsite in the lower Galilee of Israel

Thomas M. Urban a,*, Yorke M. Rowan b, Morag M. Kersel c

aResearch Laboratory for Archaeology and the History of Art, University of Oxford, Dyson Perrins Building, South Parks Road, Oxford OX1 3QY, UKbUniversity of Chicago, United StatescDePaul University, United States

a r t i c l e i n f o

Article history:Received 18 November 2013Received in revised form5 February 2014Accepted 2 March 2014

Keywords:Ground-penetrating radarIsraelChalcolithicTerra rossaLevantGeophysics

* Corresponding author.E-mail address: thomas.urban@rlaha.ox.ac.uk (T.M

http://dx.doi.org/10.1016/j.jas.2014.03.0130305-4403/� 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t

The Chalcolithic site of Marj Rabba, in the lower Galilee of Israel, features shallow limestone structuressituated in a terra rossa matrix. Calcareous substrates such as terra rossa, common throughout the region,are often not considered amenable to ground-penetrating radar (GPR) studies due to strong attenuation,particularly within the relatively high frequency range most often used in archaeological GPR surveys.Energy loss due to scattering from small embedded stones also exacerbates attenuation at this particularsite in addition to obscuring detected archaeological features, thereby complicating interpretation.Because features are fairly shallow (upper 1.5 m) and contrast well with the soil, however, GPR wassuccessful in spite of poor substrate quality. The selection of a somewhat lower antenna frequency(250 MHz) than is often recommended for archaeology, played a role in the success of the work. The endresult expands the known spatial extent of the site by five-fold, increasing our knowledge of architecturaland village plans for a time period which is poorly understood in this region. The settlement scale andcomplexity shown by these new results indicates that Chalcolithic villages are not only present in theGalilee but are as extensive and architecturally sophisticated as contemporaneous settlements in otherregions. In combination with excavation results, the structures detected with GPR at Marj Rabba providethe largest plan of an early Chalcolithic settlement in the Galilee.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

The site of Marj Rabba is a late prehistoric settlement in lowerGalilee, Israel (Fig. 1) dated to the Chalcolithic period (c. 4500e3600 BC) based on ceramic typology and radiocarbon dates. Thiskey transitional period between the Neolithic and Bronze Ageestablished new social practices, such as secondary burial in formalcemeteries, and sophisticated technological skills, perhaps mostdramatically evident in the earliest metallurgy. These indices ofincreasing social and technological complexity are complementedby a dramatic expansion in population, evident in both increasingsize and number of settlements. Our understanding of life in thesouthern Levant during this period has been largely determined bylimited survey and excavations conducted in the Jordan Valley,Negev Desert, and the Golan Heights (Rowan and Golden, 2009).Interpretations of these dramatic changes are thus based on veryfew fully published sites and a few methodical surveys. In some

. Urban).

areas, such as the Galilee, very little data is available from thisperiod; no radiocarbon dates or coherent architectural plans areavailable from another Chalcolithic settlement. The Galilee Pre-history Project, an Oriental Institute of the University of Chicagoresearch program, seeks to fill in the critical gaps of this importanttransitional period. This effort has included recent investigations atthe Chalcolithic site of Marj Rabba in the lower Galilee region(Rowan et al., 2012).

Situated on active agricultural land (Fig. 2), Marj Rabba islocated in a region of Israel that exhibits a primarily Mediterraneanclimate in accordance with standard climate classification schemes(e.g. Köttek et al., 2006). This climate makes the region moresuitable for agriculture than nearby arid areas such as the Negev ofsouthern Israel, and also fosters the development of the clay-richsoil type found at Marj Rabba. The site features circular andrectilinear stone foundations (Fig. 3) in at least three differentbuilding phases, exposed over five seasons (2009e2013) of exca-vation (Rowan and Kersel, 2014). In 2011, the Galilee PrehistoryProject team undertook non-invasive geophysical investigations atMarj Rabba in an attempt to understand the full extent of the siteand the distribution of features therein (Fig. 2). These efforts

Fig. 1. Map showing location of Marj Rabba and select Chalcolithic sites in the region.

T.M. Urban et al. / Journal of Archaeological Science 46 (2014) 96e106 97

included broad surveying with ground-penetrating radar (GPR)and magnetometry, the latter of which failed to detect architec-tural features in suitable detail. The GPR survey required specialconsideration due to the moist, clay-rich, rocky substrate at thesite as inferred from visual inspection of the soil type; a limestonederived terra rossa (Figs. 4 and 5). Conyers (2013:203e204) iden-tifies the aforementioned soil conditions (wet clay, rocky) assubstrate in which poor GPR performance can be expected in mostinstances, and previous researchers have noted poor GPR perfor-mance in calcareous substrates due to electrical losses (Grant andSchultz, 1994). In addition to this, the substrate at Marj Rabba isvery rich in iron oxides, which give the soil its characteristic redcolor. High concentrations of magnetic oxides increase the mag-netic permeability of the substrate leading to additional magnetic

losses. Though normally a much smaller factor in attenuation(Ward and Hohmann, 1987), magnetic permeability has beenshown to noticeably enhance attenuation when sufficient mag-netic material is present (Van Dam et al., 2002). Beyond theelectromagnetic loss mechanisms, the soil at Marj Rabba con-cealed shallow and variable limestone bedrock and exhibitedincoherent limestone rubble. As the architectural features at thissite are made of the same stone, an additional complicating factorwas anticipated with the work; the difficulty of distinguishingfortuitous patterns of rubble and bedrock from architecture wheninterpreting the results.

As part of the plan to mitigate the anticipated effects of atten-uation, a somewhat lower antenna frequency (250 MHz) than istypically recommended in archaeological geophysics was chosen

Fig. 2. Map of site showing locations of survey areas and excavations. The excavated portions of the site, demarcated with blue polygons, lie in between the indicated survey areas.The large survey to the north is an active olive grove in which only limited coverage was possible with GPR. (For interpretation of the references to color in this figure legend, thereader is referred to the web version of this article.).

T.M. Urban et al. / Journal of Archaeological Science 46 (2014) 96e10698

because it has long been known that attenuation is more pro-nounced at higher frequencies (Davis and Annan, 1989; Jol, 1995;Smith and Jol, 1995; Leucci, 2008). With archaeological GPR, how-ever, higher frequencies are often used in order to achieve a degreeof resolution acceptable for archaeological interpretation. Forexample, Lawrence Conyers, the most widely cited source onarchaeological GPR, has used antenna frequencies of 400 MHz andhigher for most applications and states that he has found use for a270 MHz antenna in less than 5% of circumstances (Conyers, 2012:27). He has also previously argued that with frequencies below300 MHz, resolution suitable for archaeology is unlikely (Conyers,2006: 144).

Fig. 3. Excavated Chalcolithic structure and bedrock at Marj Rabba (sq. B1, looking south).ample exposed bedrock.

Here we offer an example of a successful survey with a 250 MHzantenna in a setting where a higher frequency may have failed todetect features buried in some instances by a meter of clay-richterra rossada common substrate in the Mediterraneanbasinddemonstrating that though attenuation may limit probingdepth (especially with higher antenna frequencies) it may alsoimprove resolution (especially with lower antenna frequencies).The GPR investigation at Marj Rabba revealed many features at aresolution suitable for archaeological interpretation and for the fulldepth range in which features occurred (1.5 m), expanding theknown boundaries of the site in three directions, including into anactive olive grove to the north of current excavations. This was done

Here we see the limestone building material embedded in a clay-rich terra rossa with

Fig. 4. The olive grove north of the excavation. The photo demonstrates the rocky nature of the substrate, likely responsible for additional scattering losses and cluttering.

T.M. Urban et al. / Journal of Archaeological Science 46 (2014) 96e106 99

in a fraction of the time that would have been possible by extensiveexcavation, confirming a sprawling settlement size previously hy-pothesized from artifact distributions across the site surface. Theseresults have since been confirmed by an intensive pedestriansurvey.

2. Methods

2.1. GPR survey

The instrument chosen for the Marj Rabba GPR survey was aNoggin series common-offset GPR system by Sensors and SoftwareInc. The instrument was deployed with a monostatic 250 MHzcenter-frequency antenna (bandwidth 125e375 MHz) on a sledconfiguration, with in-line distances recorded by both a calibrated

Fig. 5. The exposed surface of the dirt road leading into the site offers a clear

odometer and a survey tape. For all data collection, a high stackingrate (8�) was chosen and a high gain setting was also selected forimproved real-time viewing.

Data collection proceeded by breaking up the site into severalareas and developing a survey strategy appropriate for each sector.Some of these surveys varied in layout and transect spacingdepending on the size of the area to be surveyed and obstaclespresent in the survey area.

Just to the north of the existing excavation area is a large olivegrove that is still in use. A gridded reconnaissance survey in thisspace was dictated by the layout of the olive trees within the grove.The grid was very coarse, with approximately 5.0 m intervals be-tween transects, which were collected in both the north/south andeast/west directions in a cross-hatch pattern. The grid layout wasdictated by the sprawling size of the area to be searched, the

glimpse of clay-rich terra rossa substrate that forms the matrix of the site.

T.M. Urban et al. / Journal of Archaeological Science 46 (2014) 96e106100

presence of the olive trees laid out in rows, and the expected scaleof archaeological features (based on previously excavated featuresat the site). This approachwas used to assess whether or not similarfeatures to those already excavated were likely to be present in theolive grove, not to map those features in high, plan-view resolution.With this simple approach, it was hoped and later shown to becorrect, that the contrast between the attenuating substrate andconcentrations of stone associated with archaeological featureswould be indefinable in profile view.

South and east of the existing excavation were large open fields(i.e. lacking the obstacles of the olive grove). These were surveyedunidirectionally (east/west strike) with a transect spacing of 0.5 m.This resolution was chosen to optimize time available (these arefairly large search areas) and data density (sufficient to detectarchitectural features in good detail). Unlike with the olive grovesurvey, the goal here was to generate good plan-view images ofarchitectural features.

Finally, a feature detected in the coarse reconnaissance survey ofthe olive grove was later surveyed bi-directionally at a transectspacing of 0.2 m. This was to provide higher resolution imaging ofthe feature, thus demonstrating the usefulness of the olive grovestrategy by providing supporting evidence that the featuresdetected in the low resolution survey are indeed archaeological.

GPR signal velocities were assessed using the technique of hy-perbola fitting, possible through the many diffraction hyperbolasfound throughout the survey area and at varying depths. Thoughvariable (ranging from 0.06 to 0.12 m/ns as discussed later), anaverage velocity of 0.07 m/ns was established and used to convertof two-way travel times into depth estimates. A variety of otherpost-processing procedures were undertaken to maximize theusefulness of the GPR survey data. These were implemented withsoftware products by Sensors and Software Inc., and included anSEC-2 gain, dewow, background subtraction, enveloping, and ve-locity migration. No topographic corrections were undertakenbecause variability in surface relief was limited at this site.Collected profiles were gridded and time-sliced at 0.2 m intervals,and Hierarchical Data Format (HDF) files generated for 3-D volumerendering. Final figures were produced with Surfer 12 and Voxler 3,both by Golden Software Inc. For a detailed discussion of the dataprocessing procedures described here see Jol and Bristow 2003,Annan 2004, Daniels 2004, Jol 2009, Conyers 2013, Goodman andPiro 2013.

2.2. Magnetic survey

A magnetic survey was conducted in the south and east fieldswith a G-858 alkali vapor magnetometer by Geometrics Inc. Theinstrument was deployed in vertical gradiometry configurationwith a 1.0 m separation between sensors. Transects were collectedanti-parallel with 0.5 m transect interval. Processing includeddespiking and dropout removal, and gridding with the Kriging al-gorithm. Final figures were produced with Surfer 12.

3. Results

The GPR surveys at Marj Rabba revealed a number of potentialarchaeological features in each survey area, thereby expanding theknown bounds of the site (Fig. 6). Most features detected appear tobe architectural remains, which are likely related to the excavatedChalcolithic structures which comprise the only ancient architec-ture found at the site to date. An overview of the GPR results isgiven below in two sections, the first focused on the olive grove andthe second focused on the south and east fields. A third sectionoffers a brief overview of themagnetic results with consideration ofhow they may relate to the GPR results.

3.1. Olive grove

The unusual survey strategy implemented in the olive groveyielded some interesting results. While unable to delineate goodplan-view architecture due to the wide transect interval necessi-tated by the grove layout, areas of potential archaeological depositswere nonetheless identified in amore binary fashion. That is, whereclusters of GPR reflections occurred within the likely depth range ofarchaeological features (upper 1.5 m) they were assumed to bepotential archaeological features. These were easily decoupledfrom the more discrete, hyperbolic anomalies generated by treeroots and in some instance overhanging limbs from the olive trees(Fig. 7). One feature identified with this survey was followed upwith a denser gridded survey shown in Fig. 8, and appears to havedetected the corner of a structure in excellent resolution The resultof this survey supported the interpretation of the coarse survey,indicating that the coarse approach was suitable to locate archi-tectural features, though without offering the detail of the densergridded surveys.

3.2. South and East fields

The south field exhibited perhaps the most exciting results ofthe GPR investigations at Marj Rabba (Fig. 9). In this survey anumber of subsurface features appeared in the upper 1.5 m. Thesefeatures appear unambiguously anthropogenic. Moreover, the scaleand layout of these features are clearly architectural and in keepingwith the known geometry of excavated features at this site and atother Chalcolithic sites in the region (Fig. 10). The features alsoindicate some variability with depth, as shown in the GPR depthslices (Fig. 9), perhaps due to multiple building phases as identifiedduring excavation at the site. The features appear to be a densecomplex of multiple walls and rooms, some possible silo bases, andat least one broad room with courtyard.

While not exhibiting the clarity of the south field results, theeast field nonetheless exhibited evidence of many sub-surfacestructural remains (Fig. 11). The disparity in the two results maystem from the fact that the east field appears to be rockier than theSouth field and may therefore exhibit more cluttering to distort theembedded archaeological features. Features that do show upclearly in the east field include many small circular features(possible silo bases), similar to some features seen in the excavation(Fig. 12). As with the south field, the east field exhibits variation infeatures with depth, probably the result of multiple building pha-ses. Line drawings and 3-D reconstructions drawn from the GPRdata for both the south and east fields are included in the figures.Close study of the GPR depth slices, however, shows a great dealmore detail than can be reasonably included in such simplifiedinterpretations. These visual aids are meant to draw attention tosome of the more prominent features and should not be taken asexhaustive interpretations.

Supplementary data related to this article can be found online athttp://dx.doi.org/10.1016/j.jas.2014.03.013.

3.3. Note on magnetic survey

The magnetic gradiometry survey conducted in the south andeast fields detected some of the same features, but with much lessdetail than the GPR results. Some insights from the magnetic data,however, may still provide useful information. Of greatest interest,the magnetic data showed a prominent linear anomaly spanningthe east/west axis of the east field (Fig. 13). Whether this anomalyhas a geologic or modern anthropogenic explanation is unknown. Ifanthropogenic and contemporary with the other features, however,this anomaly may represent a ditch, channel or other important

Fig. 6. GPR survey results. Results of the three majors survey areas are shown.

T.M. Urban et al. / Journal of Archaeological Science 46 (2014) 96e106 101

functional feature. This feature exhibits at least some spatial cor-respondence to deeper linear trend(s) in the GPR data, as seen intwo of the depth slices shown in Fig. 11 (depths exceeding 1.0 m). Inthe south field, several clusters of point dipole anomalies weredetected, indicating small magnetized sources. In contrast, very fewof these were seen in the east field magnetic result. Many of thepoint dipoles can be co-located with high amplitude GPR re-sponses. One possible explanation is that these anomalies arecaused by more recent ferrous agricultural debris. Another possible(and more tantalizing) explanation is that at least some of theseanomalies are the result of some heat intensive activity, such ascooking or pottery production, that may be related to the observedarchitectural features. Future excavationmay reveal more about themagnetic features.

Fig. 7. Olive grove survey. Low resolution gridded survey of the olive grove with an examplsubterranean archaeological feature while other diffraction hyperbolas are associated with

4. Discussion

4.1. Methodological discussion

The physical properties of the soil at Marj Rabba were qualita-tively assessed, and field strategy adjusted accordingly, on the basisof known ranges of such properties for the type of soil in question. Amore precise approach would have been to conduct an initial study,prior to GPR surveying, to assess the soil properties empirically(Doolittle and Collins, 1995). Such a study is not feasible in mostsituations, however, and the information obtained might prove tobe redundant (e.g. if you already know the matrix is moist clay) andmay therefore not be a good use of time or resources. When timeand instrument availability are limited (as is often the case) caution

e of 2 profiles collected in the Olive Grove indicating the disturbance associated with adisturbances from trees (as indicated by very high velocities).

Fig. 8. Dense survey over feature. A dense GPR survey of the same feature identified in the profiles in Fig. 7 as part of the original low resolution survey (top), indicates that clustersof reflections detected with the low resolution survey are likely caused by architectural features such as the apparent structure shown here in 3-D (bottom).

T.M. Urban et al. / Journal of Archaeological Science 46 (2014) 96e106102

can be exercised by implementing a strategy such as that describedhere (e.g. conservative antenna selection) in order to maximize thelikelihood of obtaining a useful result. Poor GPR performance wasanticipated at Marj Rabba due to knowledge of the soil type. It hasbeen argued for decades that clay can have a negative impact onGPR performance, with early researchers even suggesting that aclay content of only 5e10% (a much lower fraction than typical terrarossa) could limit probing depth to less than 1.0 m in some in-stances (Walther et al., 1986), and with more recent researchersalso suggesting that probing depth could be limited to less than1.0 m in calcareous substrates (Conyers, 2013: 56). The GPR surveyat Marj Rabba detected the full depth range of cultural features asshown in Figs. 7, 9 and 11 (up to 1.5 m), likely owing to the use of alower antenna frequency.

Velocity estimates of numerous diffraction hyperbolasthroughout the survey areas revealed two velocity trends within

Fig. 9. South field. In the south field survey area a number of features were delineated throoms and walls.

the site. First, where embedded archaeological features did notoccur or were sparse, subsurface velocities (determined bydiffraction hyperbolas generated by embedded stones) were as lowas 0.06 m/ns. This is typical for soil with a high fraction of claybecause of increased water content (Saarenketo, 1998) which maybe retained even in dry seasons. This is also consistent with visualand tactile inspection of the soil which appeared to be very moist.Within clusters of archaeological features, however, velocitiesincreased to 0.1e0.12 m/ns, indicating an abrupt vertical change inthe permittivity of the material (the transition from clay to lime-stone). This situation highlights the sharp contrast between theelectrical properties of the terra rossa and the limestone featurestherein, with these varying velocities fall within the known rangesof clay and limestone (Table 1). It is possible that if the substrate atMarj Rabba did not exhibit such a contrast, features may have beenmore difficult to detect. A similar hypothesis has been proposed by

at clearly represent subsurface architectural remains, including a number of apparent

Fig. 10. Excavation area. The excavation immediately north of the South Field GPR survey offers some clue as to the types of architectural features detected with the GPR.

T.M. Urban et al. / Journal of Archaeological Science 46 (2014) 96e106 103

other researchers working on prehistoric sites in broader region toexplain the apparent failure of GPR to detect known features(Witten et al., 2000, 2003).

Though adequate horizontal resolution was achieved with the0.5 m transect interval surveys of the south and east fields, thetransect interval probably could have been decreased by half ormore to improve resolution even further, due to shortened wave-lengths in the attenuating substrate (Appendix A). This wasdemonstrated with the small survey area in the olive grove where atransect interval of 0.2 m was used, and a correspondingly greaterlevel of deal of detail was apparent (Fig. 8). Some studies (Neubaueret al., 2002; Jol and Bristow, 2003; Novo et al., 2008) have shownclose transect spacing to significantly improve horizontal resolu-tion in GPR surveying. We believe that at Marj Rabba, transectspacing was a greater factor in determining horizontal resolution

Fig. 11. East field. Multiple circular features seen in

than was the frequency of the antenna (Appendix A) and further,that a higher frequency antenna may have failed to achieve thepenetration necessary to detect many of the archaeological featuresat the site.

4.2. Archaeological discussion

The GPR findings at Marj Rabba support the inferred size of thesitebasedon the results of themethodical pedestrian surface survey,and further expands the identification of dense architecture beyondthe boundaries of the excavated areas. This supports our recent es-timates thatMarj Rabba extended over approximately 4.5 ha, one ofthe largest known Chalcolithic sites in the western Galilee (seeShalem, 2008 forother site).More importantly, thenewfindings addto the growing body of Chalcolithic settlements in the southern

the east field may be the foundations of silos.

Fig. 12. Circular features. This circular foundation revealed in the excavation exhibits comparable scale to many circular features identified in the GPR survey, particularly in the eastfield. Such features may be the bases of silos.

Fig. 13. A comparison of magnetic (bottom) and GPR results (top). While the magnetic survey failed to detect much of the complex detail seen in the GPR results, useful comparativedata was still gained that may still offer some insights.

T.M. Urban et al. / Journal of Archaeological Science 46 (2014) 96e106104

Table 1Properties of limestone and clay. After Davis and Annan (1989). The velocity givenfor clay is at the lower end of the range and could bemore comparable to the velocityof limestone if the substrate were very dry.

Material Relativepermittivity

Conductivity(mS/m)

Velocity (m/ns) Attenuation (dB/m)

Air 1 0 0.3 0Limestone 4e8 0.5e2 0.12 0.4e1Clay 4e40 2e1000 0.06 1e300

T.M. Urban et al. / Journal of Archaeological Science 46 (2014) 96e106 105

Levant, particularly in the Galilee where few excavations have beenconducted. Connecting the South FieldGPR survey inparticularwiththe excavations, the continuation of well-built stone walls of mul-tiple rectilinear rooms and courtyards suggests that the plentifulsurface finds are a reliable indicator of the subsurface architecture.Circular features appearing in the south and east fields, though ingreater numbers in the latter, match the scale of known circularfeatures from the excavation. Thesemay be the bases of silos ormayhave served some other purpose. Though functional claims may bedifficult tomake solely on the basis of GPR data, thefindings arewellsupported with the pedestrian surface survey and excavation, andfurther provide a basis for hypotheses that may be testable throughfuture excavation. The scale, layout and type of architecture knownthrough excavation at the site indeed matches features detected inthe unexcavated areas with GPR. These results demonstrate thatChalcolithic sites were as substantial and architecturally sophisti-cated as those from other regions of similar date, confirming earliersuspicions that the lacunae of Chalcolithic sites in the Galilee mightreflect a lack of research focus, not the ancient reality (Gilead,1989).The scale and complexity of architecture exhibited in the GPR find-ings at Marj Rabba therefore indicate that such extensive villageswere not limited solely to the Jordan Valley and northernNegev, andthat any broader interpretations about Chalcolithic socio-economiccomplexity must include data from the Galilee. If the GPR resultspresented here are to be taken as an extension of the excavatedfeatures and surface artifact distribution atMarj Rabba, then theGPRresults coupled with the excavation, constitutes the largest Chal-colithic village plan yet described for the Galilee region.

5. Conclusions

This article has provided an overview of a recent GPR investi-gation at Marj Rabba, Israel, outlining an approach for the use ofGPR which could be applied at similar settings throughout the re-gion. We have made an argument for using a lower antenna fre-quency than is commonly recommended at sites with similarsubstrate conditions to Marj Rabba, and for using fairly closetransect spacing if plan view architectural remains are to bedelineated in good detail. We have also made a case for using GPRas a coarse reconnaissance method where dense data collection isnot possible or practical, as demonstrated in the olive grove at MarjRabba. Archaeologically, the research has demonstrated that sub-stantial Chalcolithic sites indeed occur in the Galilee region, andthat such sites may be suitable for investigation with geophysicalmethods.

Acknowledgments

Support for this project was provided by the Oriental Institute ofthe University of Chicago, and the University Research Council ofDePaul University. The production of this article was supported bythe Weidenfeld Research Fellowship of the Institute for StrategicDialogue. We also thank Max Price of Harvard University forassisting with data collection in the field.

Appendix A. Resolution

To demonstrate the relationship of resolution and wavelength,consider the general case with the following wave equation (Alison,2006), where f is the probing field and c is the phase velocity,

V2f� 1c2

v2f

vt2¼ 0

From this the wavelength l can be determines as l ¼ 2pc/u,where u is the angular frequency, and where spatial resolution onthe basis of phase is limited to a fraction of l/(2p). From this itfollows that decreasing l (or increasing f) will increase spatialresolution.

Maximum achievable resolution in GPR surveys is thereforelargely a functionwavelength, which is in turn a function of velocityV and operating frequency f which can be described as,

l ¼ Vf

¼ cn$f

Where n is the refraction index. Now consider that

V ¼ cffiffiffiffiffiffiffiffiffimrεr

p ¼ cn

thus velocity is determined by permittivity εr (mr typically makesonly a minor contribution). The permittivity of the soil at MarjRabba, as determined from velocity estimates would thereforeresult in correspondingly shortened wavelengths and improvedresolution.

Resolution should be considered in both the horizontal andvertical planes. Vertical resolution Dv must first be considered as afunction of time which can be described as

DvzspV4

¼ spc4

ffiffiffiffiεr

p

where sp is the pulse duration and V is velocity (Annan, 2004).When subsurface velocity is known, and wavelength in the me-dium can therefore be determined, vertical resolution can be esti-mated spatially as a fraction of l, with the common convention ofestimating vertical resolution as l/4 and a more conservative esti-mate as l/2. ). It has been suggested, however, that with excellentdata quality a vertical resolution a great as l/8 could be possible(Widess, 1973). In estimating vertical resolution at Marj Rabba wetherefore used the center frequency of 250 MHz and the estimatedaverage velocity (determined through hyperbola fitting) to firstdetermine wavelength, which we estimated to average 32 cm.Therefore the maximum possible vertical resolution that could beachieved was determined to be 0.04 m (l/8), with the more con-servative and likely estimate of 0.16 m (l/2). This is certainly withinthe range of detecting architectural ruins in reasonably good ver-tical detail.

Horizontal resolution Dh, on the other hand, can always betreated in terms of space as the dimensions of the horizontal planeare always known. There have been several conventions fordescribing horizontal resolution which rely on slightly differingcriteria (Rial et al., 2007). Some researchers, for example, havesuggested determining Dh based on the radius r of the illuminationarea as

Dh ¼ 2r where

T.M. Urban et al. / Journal of Archaeological Science 46 (2014) 96e106106

r ¼ l

4þ zffiffiffiffiffiffiffiffiffiffiffiffiffi

εr þ 1p

and z is the distance between source and target (e.g. Conyers andGoodman, 1997). With this we calculate, using a wavelength of32 cm and εr of 16, a resolution range of 24e64 cm by varying zfrom 0.2 to 1.0 m. This approach assumes a low loss condition inwhich V is controlled primarily by εr. Other methods for assessingDh may yield slightly different results. For example, a resolution of0.14e0.32 m was determined for the same range of z using

Dh ¼ 4z

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiln 2

ð2þ azÞ

s

from Daniels (2004) with a being 100 dB/m (in the lower third ofthe range from Table 1). The above favors total attenuation overεr,and indicates that spatial resolution improves as attenuation in-creases. While these approaches for approximating Dh favordifferent criteria and therefore yield varying results, it should beimmediately obvious that shallower features in media exhibitingmoderate or strong attenuation can be resolved at a level of detailthat is sometimes limited more by transect interval than operatingfrequency. That is to say that increasing antenna frequency as astrategy for improving horizontal resolution makes no sense unlesscorrespondingly smaller transect intervals (which are oftenimpractical in terms of time available and fieldwork context) arealso introduced.

The above approximations of horizontal resolution, of course,will not necessarily predict achievable resolution under real fieldconditions as many variables can change resolution for better orworse (e.g. noisiness of signal, physical properties of targets, di-mensions and orientation of features, etc.). In practice, the safestapproach is to collect the densest data practical within the allottedtime and in keeping with the goals of the survey.

Appendix B. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jas.2014.03.013.

References

Alison, W., 2006. Fundamental Physics for Probing and ImagingOxford UniversityPress, Oxford, UK.

Annan, A.P., 2004. Ground Penetrating Radar Principles, Procedures and Applica-tions Practical Processing of GPR DataSensors and Software, Mississauga,Ontario.

Conyers, L.B., Goodman, D., 1997. Ground-penetrating Radar. an Introduction forArchaeologistsAltamira Press, Walnut Creek.

Conyers, L.B., 2006. Ground-penetrating radar. In: Johnson, J.K. (Ed.), RemoteSensing in Archaeology: a North American Perspective. The University of Ala-bama Press, Tuscaloosa, AL, pp. 131e160.

Conyers, L.B., 2012. Ground-penetrating Radar Interpretation for ArchaeologistsLeftCoast Press, Walnut Creek.

Conyers, L.B., 2013. Ground-Penetrating Radar for Archaeology, third ed.AltamiraPress, Lanhan, MD.

Daniels, D.J., 2004. Ground Penetrating Radar, second ed.The Institute of ElectricalEngineers, London.

Davis, J.L., Annan, J.P., 1989. Ground-penetrating radar for high-resolution mappingof soil and rock stratigraphy. Geophys. Prospect. 37 (5), 531e551. http://dx.doi.org/10.1111/j.1365-2478.1989.tb02221.x.

Doolittle, J.A., Collins, M.E., 1995. Use of soil information to determine application ofground-penetrating radar. J. Appl. Geophys. 33, 101e108. http://dx.doi.org/10.1016/0926-9851(95)90033-0.

Gilead, I., 1989. Chalcolithic sites in Beit Netofa Valley, Lower Galilee, Israel. Pale-orient 15 (1), 263e267.

Goodman, D., Piro, S., 2013. GPR Remote Sensing in Archaeology. Geotechnologiesand the Environment, vol. 9. Springer Science, New York.

Grant, J.A., Schultz, P.H., 1994. Erosion of ejecta at Meteor Crater: constraints fromground penetrating radar. In: Proceedings Fifth International Conference onGround-Penetrating Radar. Waterloo Centre for Groundwater Research and theCanadian Geotechnical Society, Kitchener, Ontario, Canada, pp. 789e803. June12e14, 1994.

Jol, H.M., 1995. Ground penetrating radar antenna frequencies and transmitterpowers compared for penetration depth, resolution and reflection continuity.Geophys. Prospect. 43 (5), 693e709. http://dx.doi.org/10.1111/j.1365-2478.1995.tb00275.

Jol, H.M., Bristow, C.S., 2003. GPR in sediments: advice on data collection, basicprocessing and interpretations, a good practice guide. In: Bristow, C.S., Jol, H.M.(Eds.), Ground Penetrating Radar in Sediments, Geological Society SpecialPublication, vol. 211. Geological Society, London, pp. 9e27.

Jol, H.M. (Ed.), 2009. Ground Penetrating Radar (GPR) Theory and Applications.Elsevier Science, Amsterdam.

Köttek, M., Grieser, J., Beck, C., Rudolf, B., Rubel, F., 2006. World map of the KÖppen-Geiger climate classification updated. Meteorol. Z. 15 (3), 259e263.

Leucci, G., 2008. Ground penetrating radar: the electromagnetic signal attenuationand maximum penetration depth. Sch. Exch.. http://dx.doi.org/10.3814/2008/926091, 2008.

Neubauer, W., Eder-Hinterleitner, A., Seren, S., Melichar, P., 2002. Georadar in theRoman civil town Carnuntum, Austria: an approach for archaeological inter-pretation of GPR data. Archaeol. Prospect. 9 (3), 135e156. http://dx.doi.org/10.1002/arp.183.

Novo, A., Grasmeuck, M., Viggiano, D.A., Lorenzo, H., 2008. 3D GPR in archeology:what can be gained from dense data acquisition and processing?. In: 12th In-ternational Conference on Ground Penetrating Radar, June 16e19, 2008, Bir-mingham, UK.

Rial, F.I., Pereira, M., Lorenzo, H., Arias, P., Novo, A., 2007. Vertical and horizontalresolution of GPR bow-tie antennas. Advanced Ground Penetrating Radar, 2007.In: 4th International Workshop on IEEE Proceedings. Une 27e29, 2007. DOI:10.1109/AGPR.2007.386549.

Rowan, Y.M., Golden, J., 2009. The Chalcolithic period in the southern Levant: asynthetic review. J. World Prehist. 22, 1e92. http://dx.doi.org/10.1007/s10963-009-9016-4.

Rowan, Y.M., Urban, T.M., Kersel, M.M., 2012. Recent ground penetrating radardiscoveries at Marj Rabba, Israel. Antiqutiy 86 (334) (December). http://antiquity.ac.uk/projgall/rowan334/ (accessed 18.10.13.).

Rowan, Y.M., Kersel, M.M., 2014. New perspectives on the Chalcolithic Period in theGalilee: investigations at the site of Marj Rabba. In: Spencer, J.R., Brody, A.,Mullins, R. (Eds.), Festschrift in Honor of Seymour Gitin. Eisenbrauns.

Saarenketo, T., 1998. Electrical properties of water in clay and silty soils. J. Appl.Geophys. 40, 73e88. http://dx.doi.org/10.1016/S0926-9851(98)00017-2.

Shalem, D., 2008. The upper and lower Galilee in the Late Chalcolithic period. In:Bar, S. (Ed.), In the Hill-Country, and in the Shephelah, and in the Arabah(Joshua 12, 8). Studies and Researches Presented to Adam Zertal in the ThirtiethAnniversary of the Manasseh Hill-Country Survey: 96e107. Jerusalem.

Smith, D.G., Jol, H.M., 1995. Ground-penetrating radar: antenna frequencies andmaximum probable depths of penetration in Quaternary sediments. J. Appl.Geophys. 33, 93e100.

Van Dam, R.L., Van Den Berg, E.H., Van Heteren, S., Jeroen, K.C., Kenter, A.M.,Groen, K., 2002. Influence of organic matter in soils on radar-wave reflection:sedimentological implications. J. Sediment. Res. 72, 341e352.

Walther, E.G., Pitchford, A.M., Olhoeft, G.R., 1986. A strategy for detecting subsurfaceorganic contaminants. In: Natl. Water Well Assoc./API Conf. Petrol. Hydrocarb.Org. Chem. Ground Water Prev., Detect. Restor., Houston, Nov. 12e14. Natl.Water Well Assoc, Dublin, OH, pp. 357e381.

Ward, S.H., Hohmann, G.W., 1987. Electromagnetic theory for geophysical applica-tions (Ch. 4). In: Nabighian, M.N. (Ed.), I31eI308 in Electromagnetic Methods inApplied Geophysics. vol. 1, Theory, Investigations in Geophysics Series, No. 3.Society of Exploration Geophysicists, Tulsa, OK.

Widess, M.B., 1973. How thin is this bed? Geophysics 38, 1176e1180.Witten, A.J., Calvert, G., Witten, B., Levy, T.E., 2003. Magnetic and electromagnetic

induction studies at archaeological sites in southwestern Jordan. J. Environ. Eng.Geophys. 8 (3), 209e215.

Witten, A.J., Levy, T.E., Adams, R.B., Won, I.J., 2000. Geophysical surveys in JebelHamrat Fidan, Jordan. Geoarchaeology 15, 135e150. DOI: 10.1002/(SICI)1520-6548(200002)15:2<135::AID-GEA2>3.0.CO;2-M.