causes of ground-penetrating radar reﬂections in sedimentrvd/pub/rvddissertation.pdf · causes of...

Causes of

ground-penetrating radar

reflections in sediment

Remke L. van Dam

The research reported in this thesis was carried out at the:

Vrije UniversiteitFaculty of Earth SciencesDe Boelelaan 10851081 HV AmsterdamThe Netherlands

Support for the research reported in this thesis was provided by:

Netherlands Institute of Applied Geoscience, TNO-NITG — Financial support, GPR equipmentVrije Universiteit Industrial Associates in Sedimentology — Financial supportDelft University of Technology — GPR EquipmentUniversity of Amsterdam — TDR Equipment

Cover design: Gabriela de Aguiar

c© Copyright 2001, R.L. van Dam, [email protected]

ISBN 90 9015256 3

VRIJE UNIVERSITEIT

Causes of

ground-penetrating radar

reflections in sediment

ACADEMISCH PROEFSCHRIFT

ter verkrijging van de graad van doctor aande Vrije Universiteit Amsterdam,op gezag van de rector magnificus

prof.dr. T. Sminia,in het openbaar te verdedigen

ten overstaan van de promotiecommissievan de faculteit der Aardwetenschappen

op maandag 17 december 2001 om 13.45 uurin het hoofdgebouw van de universiteit,

De Boelelaan 1105

door

Remke Leander van Dam

geboren te Hoogland

Promotor: prof.dr. W. Schlager

To my parents

”Literarum radices amaras, fructus dulces”

— Cicero —

Contents

Summary 1

Samenvatting 3

1 General introduction 51.1 Goal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51.2 Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 Identifying causes of radar reflections using time-domain reflectometry and sedimentologi-cal analyses 112.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.2 Electromagnetic wave theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2.3.1 Ground-penetrating radar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.3.2 Time-domain reflectometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162.3.3 Sedimentological and textural characteristics . . . . . . . . . . . . . . . . . . . 172.3.4 Synthetic radar traces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.4 Sedimentology and stratigraphy of the study site . . . . . . . . . . . . . . . . . . . . . . 172.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.5.1 Ground-penetrating radar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.5.2 Textural characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222.5.3 Time-domain reflectometry and lacquer peels . . . . . . . . . . . . . . . . . . . 232.5.4 Synthetic radar traces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

2.6 Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3 Iron oxides as a cause of GPR reflections 293.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.3.1 Time-domain reflectometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.3.2 Thermogravimetric analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323.3.3 Magnetic measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323.3.4 GPR modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

3.4 Sampling and sediment description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.4.1 Sampling and TDR tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.4.2 Thermogravimetric analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

3.5 Results - electromagnetic properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383.5.1 Magnetic measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383.5.2 TDR measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

3.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

Contents

3.6.1 Iron-oxide material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.6.2 Volumetric water content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413.6.3 GPR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

4 Influence of organic matter in soils on radar-wave reflection 474.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

4.2.1 Ground-penetrating radar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484.2.2 Time-domain reflectometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 504.2.3 Water-retention characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.3 Sedimentology and stratigraphy of the test site . . . . . . . . . . . . . . . . . . . . . . . 524.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

4.4.1 Ground-penetrating radar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564.4.2 Sediment description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564.4.3 Water-retention curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

4.5 Synthetic GPR modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 614.6 Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

5 Radar reflections from sedimentary structures in the vadose zone 695.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 705.2 Water retention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 725.3 Test site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

5.3.1 Experimental procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 735.3.2 Sedimentology and stratigraphy . . . . . . . . . . . . . . . . . . . . . . . . . . 765.3.3 GPR measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

5.4 TDR measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 785.5 Analysis of thin sections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

5.5.1 Image analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 795.5.2 Estimated small-scale water-retention characteristics . . . . . . . . . . . . . . . 825.5.3 Dielectric properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

5.6 GPR synthetic modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 855.7 Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

6 Synthesis 916.1 Petrophysical techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

6.1.1 Time-domain reflectometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 916.1.2 Water-retention characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

6.2 Causes of radar reflections - how widely applicable are the conclusions? . . . . . . . . . 936.2.1 Iron oxides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 936.2.2 Sedimentary structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

6.3 General conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

List of used symbols and acronyms 99

References 101

Acknowledgements 109

viii

Summary

Causes of ground-penetrating radar reflections in sediment

Ground-penetrating radar (GPR) is a geophysical technique that images the shallow subsurfaceby measuring differences of electromagnetic properties in sediment. Water plays an importantrole in the behavior of GPR waves in sediment because of its large electromagnetic contrastwith both sediment grains and air. Changes in water content that trigger GPR reflections arecommonly caused by variations in textural characteristics of sediment. As direct informationof the sediment from outcrops or cores is not generally available, interpretation of radar im-ages is not straightforward. Although electromagnetic wave theory is well established, there isa lack of detailed and quantitative understanding of radar-wave reflections in sediment. Thisthesis aims to improve understanding of the causes of ground-penetrating radar (GPR) reflec-tions in sediment by using petrophysics to describe and quantify sediment characteristics andelectromagnetic properties. Focus is on unsaturated (vadose-zone) sediments that contain threephases of matter (solid grains, water and air), which is the common situation for most GPRsurveys. Case studies were performed at two sites in The Netherlands with relatively uniformeolian sandy sediment. GPR surveys at different frequencies (25 to 900 MHz) are combinedwith petrophysical measurements of electromagnetic properties, water-retention characteristics,and texture from quarry and trench walls.

In Chapter 2, time-domain reflectometry (TDR) is introduced as a method to conduct de-tailed measurements of electromagnetic wave velocity in sediment along vertical transects inthe subsurface. It is shown that excursions in electromagnetic wave velocity, which are su-perimposed on a baseline value for dry sand, are caused by the presence of organic materialand iron-oxide bands. Sedimentary structures lead to clear excursions only when the porescontain some water. Synthetic radar traces, which were constructed using the TDR logs andsedimentological information, show that radar results may be ambiguous because of multiplesand interference. The results indicate that water content is the most important factor affectingelectromagnetic properties of sediment. Consequently, the ability of the sediment to hold watergoverns GPR reflections.

Iron oxides frequently occur as secondary precipitates in sediment and may form bands orirregular patterns. In chapter 3, TDR field studies show that goethite, which is the most abundantiron-oxide precipitate in soils around the world, significantly lowers the electromagnetic wavevelocity of sediment. For a set of samples containing a varying amount of precipitates, magneticmeasurements and laboratory TDR tests demonstrate that goethite does not affect magneticor dielectric properties of sediment directly. However, the variation in electromagnetic wavevelocity is related to the amount of capillary water in the sediment, differences in which arecaused by a higher water-retention capacity of goethite relative to quartz grains. Iron oxidesthus can have a profound influence on the water content and, consequently, on the GPR signal.

Summary

Using 2-D synthetic radar sections it is shown that patterns of iron-oxide precipitation, whichare governed by sedimentary structures, discontinuities, and groundwater flow paths, influencethe GPR reflection configuration and can cause major difficulties in interpretation.

Soils are excellent reflectors of GPR signals, because of their ability to hold water. Inchapter 4, GPR profiles of an eolian sedimentary succession and measurements of texture, di-electric properties, and water retention demonstrate the effect of water concentrated in soils onradar-wave reflection. Water-retention curves, which plot water content with change in suctionpotential, depend on the organic-matter content of sediment. As a result of a uniform pore-sizedistribution, clean sand and weakly developed soils experience a rapid transition from saturatedto almost dry conditions. In contrast, prominent soils show a more gradual decrease in watercontent with increasing suction. The variations in curve shape imply that contrasts in dielec-tric properties between adjacent units will differ according to the moisture conditions. Whenclean sand drains, the reflection coefficient between clean sand and a prominent soil increasesabruptly, reaches a maximum value at field-capacity conditions, and then decreases slowly.Synthetic GPR images show that field-capacity conditions, with high reflection coefficients, arefavorable for tracing one single soil horizon. When focusing on stacked soils, saturated or fullydry conditions are preferable.

Although GPR is a useful technique to image sedimentary structures in the vadose zone,it is not known in great detail which textural characteristics control water retention and, thus,reflections. Also, little is known about the contribution of structures smaller than the resolutionof the radar wave to the total reflection. To address these issues, Chapter 5 combines GPRimages of high-angle eolian foresets with petrophysical measurements. Water content in thevadose zone is governed by the characteristics of the pore network, which can be approximatedby a combination of grain-size distribution and total porosity. Under unsaturated conditions,fine-grained and tightly packed material retains more water than coarse and loosely packed ma-terial. TDR measurements indicate that small texture variations in a 5-cm-thick foreset causeclear GPR reflections. In an experimental approach to model the radar-wave response of in-dividual wave trains across structures that are finer than the resolution of the GPR waves, acentimeter-scale digitized thin section was used to render detailed textural information nor-mal to the bedding. A neural network and a dielectric mixing model allowed reconstructionof water-retention characteristics and electromagnetic wave impedance. Synthetic radar tracesshow that sub-centimeter-scale layering causes interference of reflected signals, which is rec-ognized in reflection patterns that change with frequency. Although in an actual measurement,with spherical wavefront propagation, the interfering signal will be low-amplitude, reflectionsfrom sub-centimeter-scale layering may cause loss of energy and may explain reflections thatdo not represent original bedding.

Concluding, it can be said that TDR has proven very useful to quantify electromagneticproperties of sediment and to understand causes of GPR reflections. Radar reflections in sed-iment result primarily from variations in water content that cause contrasts in electromagneticproperties of adjacent layers. Variations in water content can be attributed to the presence oforganic matter or diagenetic iron oxide, preventing interstitial water from draining. Also, dif-ferences in water retention result from small texture variations in sedimentary structures.

2

Samenvatting

Oorzaken van grondradar-reflecties in sediment

Grondradar (ground-penetrating radar, GPR) is een geofysische techniek die door het meten vanveranderingen in elektromagnetische eigenschappen een beeld verkrijgt van de ondiepe onder-grond. Door het grote elektromagnetische contrast met sedimentkorrels en lucht speelt watereen belangrijke rol in het gedrag van radargolven. Veranderingen in texturele eigenschappenvan sediment leiden tot variaties en watergehalte die vervolgens radarreflecties initieren. Doorde afwezigheid van informatie over het sediment is de interpretatie van radarbeelden vaak nieteenvoudig. Hoewel de theorie van elektromagnetische golven goed bekend is, bestaat er eengebrek aan gedetailleerd en kwantitatief inzicht in de oorzaken van grondradar-reflecties in sed-iment. Dit proefschrift heeft als doel om deze lacune door het gebruik van petrofysica weg tenemen. Petrofysische technieken beschrijven en kwantificeren texturele en elektromagnetischeeigenschappen van sediment. De nadruk ligt op onverzadigd sediment, bestaande uit korrels,water en lucht, omdat dit de meest gebruikelijke situatie is bij grondradar-metingen. Op tweelokaties in Nederland zijn veldstudies uitgevoerd in eolische afzettingen met relatief uniformzand. Grondradar-metingen in het frequentiebereik tussen 25 en 900 MHz zijn gecombineerdmet petrofysische metingen van elektromagnetische eigenschappen, watergehalte en textuur.

In Hoofdstuk 2 wordt tijdsdomein-reflectometrie (TDR) geıntroduceerd voor het gedetail-leerd meten van de variatie in elektromagnetische golfsnelheid langs verticale profielen in deondergrond. Metingen in een groeve laten zien dat afwijkingen in golfsnelheid, die zijn gesuper-poneerd op een basiswaarde voor droog zand, kunnen worden gecorreleerd aan het voorkomenvan ijzerbanden en bodems. Sedimentaire structuren leiden pas tot waarneembare afwijkingenin golfsnelheid wanneer het sediment niet volledig droog is. Synthetische radarsignalen, gecon-strueerd met behulp van de TDR-metingen en textuurinformatie, laten zien dat grondradar-beelden door interferentie soms lastig te interpreteren zijn. Water is de belangrijkste factorvoor veranderingen van elektromagnetische eigenschappen van sediment. Het vermogen vansediment om water vast te houden is dus de bepalende factor in het onstaan van grondradar-reflecties.

IJzeroxides komen vaak in banden of in onregelmatige patronen voor in sediment. In Hoofd-stuk 3 tonen TDR-veldmetingen aan dat goethiet, hetgeen het meest voorkomende ijzeroxide is,de elektromagnetische golfsnelheid van sediment significant verlaagt. Magnetische- en TDR-metingen aan een serie monsters varierend in hoeveelheid aanwezig ijzeroxide, laten zien datgoethiet de magnetische of dielektrische eigenschappen van het sediment niet direct beinvloedt.De waargenomen variatie in golfsnelheid kan wel verklaard worden door de hoeveelheid capil-lair water in het sediment. Dit wordt veroorzaakt door een hoger water-vasthoudend vermogenvan ijzeroxides ten opzichte van dat van kwarts. Goethiet kan dus een significante invloed uitoe-fenen op het watergehalte, en daarmee op het radarsignaal. Twee-dimensionale radarprofielen

Samenvatting

laten zien dat het patroon van ijzeroxide neerslag, dat wordt beinvloed door sedimentaire struc-turen, onregelmatigheden en de grondwaterstroming, problemen kan veroorzaken bij het inter-preteren van grondradar-beelden.

Bodems reflecteren grondradar-signalen zeer goed als gevolg van hun water-vasthoudendvermogen. In Hoofdstuk 4 demonstreren radarprofielen van een eolische afzetting en metin-gen van textuur, dielektrische eigenschappen en water-vasthoudend vermogen het effect vanbodemwater op grondradar-reflecties. Water-retentie curves die watergehalte tegen zuigspan-ning uitzetten, zijn gecorreleerd met het gehalte organische stof in het sediment. Als gevolg vaneen uniforme poriengrootteverdeling gaan schoon zand en slecht ontwikkelde bodems in eenzeer smal zuigspanningsbereik van verzadigd naar droog. Beter ontwikkelde bodems kenneneen gelijkmatiger afname. De verschillen betekenen dat contrasten in elektrische eigenschappentussen verschillende eenheden zullen varieren al naar gelang de bodemvocht condities. De re-flectiecoefficient tussen een goed ontwikkelde bodem en schoon zand neemt snel toe als schoonzand draineert. Het bereikt een maximum rond veldcondities en neemt daarna geleidelijk af.Synthetische radarbeelden geven aan dat onder veldcondities, wanneer de reflectiecoefficientgroot is, een bodem het best in beeld kan worden gebracht. Wanneer meerdere gesuperponeerdebodems in beeld moeten worden gebracht zijn verzadigde of volledig droge condities beter.

Grondradar is een goede techniek voor het in beeld brengen van sedimentaire structuren inde onverzadigde zone, maar het is niet precies bekend welke texturele eigenschappen water-retentie karakteristieken, en dus reflecties, controleren. Ook de minimaal noodzakelijke ve-randering voor een reflectie, en de bijdrage van structuren kleiner dan de golfresolutie aanhet totaalresultaat, zijn niet bekend. Hoofstuk 5 combineert grondradar-beelden van eolis-che sedimentaire structuren met petrofysische metingen. Het watergehalte in de onverzadigdezone wordt bepaald door de porienstructuur en kan worden benaderd door meting van korrelg-rootteverdeling en totale porositeit. Onder onverzadigde condities houdt fijnkorrelig en dicht-gepakt sediment meer water vast dan grof en losgepakt materiaal. TDR-metingen laten zien datkleine variaties in korrelgrootteverdelingen in een 5-cm dikke structuur duidelijke radarreflec-ties veroorzaken. In een meer experimentele benadering om de respons van radargolven overstructuren, fijner dan de maximale resolutie, te modelleren, is een gedigitaliseerd slijpplaatje ge-bruikt voor het bepalen van texturele veranderingen (¡1 cm) loodrecht op de gelaagdheid. Eenneuraal netwerk en een ’dielektrisch mixing model’ zijn vervolgens gebruikt voor het recon-strueren van het water-vasthoudend vermogen en reflectiecoefficienten. Synthetische metingenlaten zien dat de gelaagdheid leidt tot interferentie van gereflecteerde golven, hetgeen resul-teert in een reflectiepatroon dat verandert voor iedere gebruikte frequentie. In een veldmet-ing, met een golffront dat zich voortplant in 3 dimensies, zal een meer prominente reflectiedomineren over interfererende reflecties met een lagere amplitude. Echter, de reflecties van fi-jne gelaagdheid kunnen leiden tot een verlies van elektromagnetische energie en vormen, doorde interferentie, een verklaring voor reflecties die geen duidelijke laagovergangen representeren.

Concluderend kan worden gesteld dat TDR en zeer bruikbare techniek is voor het kwantifi-ceren van elektromagnetische sedimenteigenschappen en voor het begrijpen van oorzaken vangrondradar-reflecties. Reflecties zijn voornamelijk het resultaat van veranderingen in waterge-halte die leiden tot elektromagnetische contrasten in sediment. Variaties in watergehalte kunnensamenhangen met het voorkomen van organische stof en ijzeroxides, die het water vasthouden.Ook wordt het water-vasthoudend vermogen beinvloed door kleine textuurveranderingen in sed-imentaire structuren.

4

1

General introductionGround-penetrating radar (GPR) is a geophysical technique that is widely used to study theshallow subsurface in a broad range of applications and environmental settings. GPR mea-sures changes in the electromagnetic properties of subsurface features that cause reflection oftransmitted electromagnetic waves. The technique emerged in the 1940s but was first used forgeotechnical issues in the 1970s (Ulriksen, 1982; Olhoeft, 1988). From the 1980s, when the firstcommercial systems became available (Daniels et al., 1988), the use and variety of applicationsof GPR have seen an advance that has continued to the present day (Figure 1.1). The use ofGPR in sedimentological research is a more recent development. Early work was performed by,amongst others, Jol and Smith (1991) and Schenk et al. (1993). Since the middle of the 1990sthis field has seen a spectacular increase (Figure 1.1). Nevertheless, because of the relativelylow economic relevance, the developments have not progressed as quickly as in seismics.

It is generally understood that water plays a crucial role in the behavior of GPR waves (Davisand Annan, 1989). In sedimentary environments, variations in textural characteristics com-monly cause the necessary changes in water content that trigger radar-wave reflections (e.g.,Hanninen, 1992; Huggenberger, 1993). With water as a major source of environmental andsocial problems like droughts, flooding and pollution both now and in the future (Vorosmartyet al., 2000) there is a growing need for accurate techniques like GPR that can assess water-related phenomena in the subsurface. To the present date, GPR studies with earth-science ob-jectives have commonly focused on geological and hydrological reconnaissance or on syntheticmodeling of wave propagation, rather than on a combination of the two. Studies integratingdetailed information of sediment properties with modeling of electromagnetic-wave propaga-tion and reflection to understand GPR reflections in sediment are sparse. As a result, littlequantitative knowledge exists on the actual causes of radar-wave reflection in sediment.

1.1 Goal

The principal goal of the thesis is to understand the origin of ground-penetrating radar (GPR) re-flections in sediment, using case studies from The Netherlands. This subject is interdisciplinaryand can be positioned between research fields of sedimentology, petrophysics, and syntheticmodeling (Figure 1.2). GPR studies with geological objectives commonly use the radar imagesto understand sedimentary features in the shallow subsurface, but simultaneously, sedimento-logical concepts are the key to explanation of the images. As direct information about thesediment (from cores or nearby outcrops) is generally sparse, image interpretation of GPR datais often difficult. Geophysical signal analysis may help with interpretation but GPR-syntheticapproaches usually lack input and understanding on actual sediment properties. Due to thenon-unique interpretation of most geophysical images, a full understanding and adequate inter-

Chapter 1

439

318

104

29

2

62

86000

100

200

300

400

500

1996-20001991-19951986-19901981-19851976-1980

Years

GPR & geology

GPR & sedimentology

Num

bers

of p

ublic

atio

ns

Figure 1.1: Increase in the number of scientific publications on and using ground-penetrating radar since1976. From the early 1980s there is a steady increase in the number of publications on the geologicalapplications of GPR. Halfway the 1990s the number of publications on sedimentological applications ofGPR increased significantly. The information is obtained from GeoRef, an American Geological Institutedatabase. Search words entered were ”GPR”, ”ground-penetrating radar”, ”ground penetrating radar”,”georadar”, ”geology”, ”sediment”, and ”sedimentology”. c©1990-2001 American Geological Institute,SilverPlatter International, N.V.

pretation of GPR sections is often impossible, even with help of signal analysis and core andoutcrop information.

Many subsurface variations that influence electromagnetic waves are beyond the resolutionof GPR. However, for a full understanding of the origin of reflections, the complete GPR sen-sitivity range should be analyzed in analogy to reflection seismics (e.g., Mayer, 1980). Petro-physics is the factor that builds a bridge between sedimentology and synthetic modeling (Fig-ure 1.2). It describes material in terms of sediment characteristics and electromagnetic proper-ties. I have three objectives in using petrophysical techniques to understand different causes ofGPR reflections in sediment:

1. establishing a method that enables quantification of electromagnetic properties in the sub-surface,

2. defining sediment characteristics that may affect GPR signals, and

3. quantifying the influence of these characteristics with respect to GPR reflections.

6

General introduction

Models ofwave

impedance

Texturalcharacteristics

Electromagneticproperties

Water-retentioncharacteristics

Organic-mattercontent

4

Iron-oxideprecipitates

3

Sediment body

GPR synthetics

Petrophysics

Sedimentology Synthetic modeling

Cha

ract

eriz

atio

n of

text

ural

pro

perti

es Characterization of w

ave impedance

Signal analysisImaging

GP

Rinterpretation

Figure 1.2: Ground-penetrating radar positioned in the center of a three-discipline triangular config-uration. The thick line that goes from GPR through sedimentology to petrophysics and via syntheticmodeling back to GPR shows the approach used in the chapters of this thesis. The flow diagram withrectangular image boxes that is positioned around the triangle shows the followed procedure and usedmethods in greater detail. The numbers in the lower-right corner of some boxes refer to the chapters ofthis thesis with emphasis on these specific situations.

7

Chapter 1

50km0

N

UtrechtKatwijk aan Zee

Amsterdam

Ossendrecht

Meuse

Scheldt

Rhine

Figure 1.3: Map of The Netherlands showing the locations of study sites.

Data were collected at two sites in The Netherlands, Ossendrecht and Katwijk aan Zee(Figure 1.3). The sediment at both sites is largely of eolian origin and has a relatively uniformgrain-size distribution with little or no clay and gravel. At these sites sedimentary structures, soilhorizons, and iron-oxide precipitates cause variations in textural and hydrological properties.My focus is on unsaturated (vadose-zone) sediments that contain three phases of matter - solidgrains, water and air. This is the common situation for most GPR surveys. Saturated sedimentis generally considered unpreferable because of low signal penetration, whereas truly dry (two-phase) systems are uncommon.

1.2 Approach

The structure of the thesis is such that all disciplines in Figure 1.2 (sedimentology, petrophysics,and synthetic modeling) are integrated in each chapter. First, I used sedimentological descrip-tions from quarry and trench walls to get an overview of the sediment units involved. Second,the units were surveyed with GPR at different frequencies. Third, I used petrophysics to con-struct vertical models of electromagnetic wave impedance. The petrophysical information usedcan be subdivided into direct measurements of electromagnetic properties, measurements ofwater-retention characteristics and measurements of textural characteristics (Figure 1.2). Thevertical models of electromagnetic wave impedance are used to construct GPR synthetic images

8

General introduction

and allow for an improved understanding of the original image or give, together with the previ-ous steps, insight into the more theoretical aspects of GPR reflection in sediment (Figure 1.2).

Four chapters (2 - 5) are presented that contribute to the three previously outlined objectives.Chapter 2 is concerned with the first two objectives and presents a review and theoretical back-ground of the GPR technique. Time-domain reflectometry (TDR) is introduced as a method toconduct detailed measurements of electromagnetic-signal velocity and electromagnetic proper-ties in the subsurface. Along vertical transects in a quarry, TDR measurements and sedimen-tological observations are used to define sediment characteristics that may affect GPR signals.These sediment characteristics can be subdivided into iron-oxide precipitates, organic-mattercontent and sedimentary structures (Figure 1.2). Synthetic modeling of GPR signals is usedto show the effects of these sediment characteristics on GPR reflection and interference. Theresults are used as a starting point for further research and quantification (third objective) in thenext chapters.

Chapter 3 focuses on the special situation of iron-oxide precipitates. TDR measurementsshowed that unsaturated sediment with precipitated goethite experiences a decrease in electro-magnetic wave velocity. Using dielectric mixing models and 2-D synthetic radar sections theimpact of iron oxides on the GPR signal and reflection configuration is shown. Chapter 4 dealsspecifically with soil horizons, which are excellent reflectors of GPR signals, owing to the abil-ity of organic matter to hold water. Measurements of water-retention characteristics are used toquantify dielectric contrasts and radar-wave reflectivity between clean sand and soils at differ-ent suction potentials. Chapter 5 deals with the observation from Chapter 2 that small texturalvariations in sedimentary structures can cause clear GPR reflections. TDR measurements anddigital images of thin sections are used to quantify texture variations in sedimentary layeringand to study the response of GPR signals from sedimentary structures. Chapter 6 synthesizesthe results by broadening the context of two used methods and by generalizing some of thepresented findings.

9

Chapter 1

10

2

Identifying causes of radar reflectionsusing time-domain reflectometry andsedimentological analyses

AbstractAlthough electromagnetic theory is well established, little is known about theexact cause of radar reflections in sediment. To obtain detailed informationon the radar-wave contrast parameter in the subsurface, 100- and 200-MHzradar surveys of eolian deposits in a quarry were combined with time-domainreflectometry (TDR). Combining lacquer peels and TDR data from the quarrywall allowed identification of relationships between sediment characteristicsand electromagnetic properties. Peaks in the product of relative permittivityand relative magnetic permeability (inversely proportional to electromagneticwave velocity) which are superimposed on a baseline value for dry sand, arecaused by potential radar reflectors. These peaks coincide with the presenceof either organic material, having a higher water content and relative permit-tivity than the surrounding sediment, or iron-oxide bands, causing water tostagnate and possibly enhancing relative magnetic permeability. Sedimen-tary structures, as reflected in textural change, only result in possible radarreflections when the pores contain capillary water. Synthetic radar traces,constructed using the TDR logs and sedimentological data from the lacquerpeels, provide an improved insight into the behavior of radar waves and showthat radar results may be ambiguous because of multiples and interference.

This chapter is based on Van Dam, R. L., and Schlager, W. (2000). Identifying causes of ground-penetratingradar reflections using time-domain reflectometry and sedimentological analyses. Sedimentology, 47(2), 435-449.Reprinted with permission.

Chapter 2

2.1 Introduction

The seismic reflection method and ground-penetrating radar (GPR) are important geophysicalreconnaissance tools for the shallow subsurface. Both techniques are based on wave propaga-tion and reflection, but wave length ranges, and consequently resolution and penetration depths,differ significantly. The seismic reflection method has a better penetration but lower resolutionthan GPR. Also, the contrast parameter for the seismic reflection method (acoustic impedance)differs from that for GPR (electromagnetic wave impedance). Yet, the two techniques are suffi-ciently analogous for experience gained with the one to be applicable to the other, increasing ourknowledge on the origin and interpretation of reflections in both the seismic reflection methodand GPR (Cardimona et al., 1998).

In the seismic reflection method, most reflections of sedimentary deposits are parallel todepositional bedding. Two different types of reflection exist (Mayer, 1980): (1) reflections thatoccur at the location of major changes in acoustic impedance; and (2) reflections that representpatterns resulting from constructive and destructive interference between the acoustic wave trainand small impedance variations in the sediments or rocks. Only the first type of reflectionserves as a reliable guide to depositional history and geometry and its amplitude is proportionalto the magnitude of change. In contrast, the second type of reflection does not represent amajor impedance change at the measured travel time (Mayer, 1979). An important method fordistinguishing these two reflection types is detailed comparison of seismic traces with sedimentand rock properties, as determined from cores and outcrop studies. These outcrop studies areconducted in deposits that are considered to be analogous, or on exposures in the vicinity of theseismic line (Stafleu and Sonnenfeld, 1994). The sediment and rock properties from the outcropstudies can be used to generate lithologic and acoustic impedance models and are widely usedto construct synthetic seismic traces and sections (Fagin, 1991; Stafleu and Schlager, 1995;Bracco Gartner and Schlager, 1999). These synthetic images improve the understanding ofseismic sections and reflections. It cannot be assumed, however, that the rock and sedimentproperties measured in outcrops are completely similar to those reflected in the seismic trace.

GPR measures changes in the electromagnetic properties of sediments that cause reflec-tion of electromagnetic energy. These changes in electromagnetic properties result primarilyfrom changes in water content, governed in turn by grain-size and porosity (Topp et al., 1980;Roth et al., 1990; Sutinen, 1992; Huggenberger, 1993). Since grain-size and porosity changesare related to depositional history, a clear relationship can be expected between sedimentarystructures and electromagnetic properties, allowing accurate identification of radar facies andsequence boundaries (Gawthorpe et al., 1993). Most work with GPR has focused either ongeological and sedimentological reconnaissance (Jol and Smith, 1991; Huggenberger, 1993;Beres et al., 1995; Bristow et al., 1996; Asprion and Aigner, 1997; Van Heteren and Van DePlassche, 1997; Bridge et al., 1998; Van Overmeeren, 1998; Neal and Roberts, 2000) or onsynthetic modeling of wave propagation (Carcione, 1996; Casper and Kung, 1996; Hollenderand Tillard, 1998). Studies in which wave theory is integrated with geological or petrophysicaldata, common in reflection seismics, are sparse. As a result, little qualitative and quantitativeknowledge exists on the cause of GPR reflections. GPR shows sedimentary structures and otherfeatures in the subsurface, but no information is available on the origin of reflections or on thebehavior of the contrast parameter, other than from the reflection image (and its parameters,such as reflection strength). However, many impedance changes are beyond the resolution andsensitivity range of GPR. For a full understanding of the origin of GPR reflections, impedancevariations smaller than those observed on radar images must be analyzed and explained.

12

Identifying causes of radar reflections

In this study, time-domain reflectometry (TDR), a technique well known in soil sciences(Topp et al., 1982; Heimovaara et al., 1995), was used to correlate GPR data with subsurfaceinformation by measuring electromagnetic properties of sediment at small intervals along ver-tical sections in a quarry. TDR measures the propagation velocity of an electromagnetic wavealong a steel-rod probe that is pushed into the sediment, and is best applied in unconsolidatedsand-sized and finer sediments. In coarse grained sediment the TDR method functions worsedue to disturbance of the sediment fabric during the penetration of the TDR rods. The TDRmeasurements were interpreted using detailed sedimentological information from lacquer peelsand grain-size analyses, providing insight into the possible causes of GPR reflections. Basedon the TDR logs and the sedimentological information from the lacquer peels, an impedancemodel of the subsurface was constructed. This model, in which the sediment column is subdi-vided into distinct layers with characteristic electromagnetic properties, was used to construct1-D synthetic radar traces. In analogy to the seismic reflection method, such synthetic imagescan be compared with real GPR sections for a better understanding and quantification of param-eters controlling GPR reflections. Ultimately, such comparisons will improve the knowledge ofwaves and their reflections in general, which may also benefit the seismic reflection method.

2.2 Electromagnetic wave theory

The techniques of ground-penetrating radar (GPR) and time-domain reflectometry (TDR) arebased on propagation and reflection of electromagnetic energy in the subsurface. The (frequen-cy-dependent) properties that control the behavior of electromagnetic energy in a mediumare dielectric permittivity (ε), where ε = ε0εr, electrical conductivity (σ), and magnetic per-meability (µ), where µ = µ0µr (Von Hippel, 1954). Here, ε0 is the permittivity of vacuum(8.85419 10−12 F m−1), εr is the relative permittivity, µ0 is the magnetic permeability of vac-uum (4π10−7 H m−1), and µr is the relative magnetic permeability. Together, these propertiesdefine the impedance (Z) of the medium for harmonic electromagnetic waves with exp( jωt)dependence (Brewster and Annan, 1994; Fokkema et al., 2001):

Z =

√jωµ

σ+ jωε, (2.1)

where j =√−1, ω = 2π f is angular frequency [radians s−1], and t is time [s]. A range of

electromagnetic properties for common geologic materials can be found in Davis and Annan(1989) and in Van Heteren et al. (1998). For most geological materials, the relative magneticpermeability (µr) is near unity (e.g., Roth et al., 1990). Consequently, the magnetic permeabil-ity in the subsurface is near the free-space value (µ0) and plays no role in the electromagneticenergy behavior (Powers, 1997). However, under certain conditions, such as the presence ofiron and iron oxides, relative magnetic permeability can be enhanced significantly (Von Hippel,1954; Olhoeft and Capron, 1994). In most natural low-loss material such as clean sand, the in-fluence of σ on the electromagnetic signal is negligible between 100 and 1000 MHz (Davis andAnnan, 1989). In contrast, the relative permittivity plays an important role in both propagationand reflection of electromagnetic waves (Huggenberger, 1993). Because water has a relativepermittivity of around 80, whereas air and quartz have values of 1 and around 4.3, respectively,the relative permittivity of sediment is governed by its water content.

13

Chapter 2

The propagation velocity (v) of electromagnetic waves in a medium is found by:

v =c0√

εrµr1+√

1+(σ/ωε)2

2

, (2.2)

where c0 is the electromagnetic wave velocity in vacuum (3 108 m s−1) and σ/ωε is a loss factor.For common geological material, the influence of the conductivity (σ) on the electromagneticwave velocity is negligible over the entire GPR frequency range (Davis and Annan, 1989) andthe loss factor approaches zero. Consequently, the velocity is mainly controlled by εr and µr

and Equation (2.2) can be simplified by:

v =c0√εrµr

, (2.3)

which reduces to:v =

c0√εr

, (2.4)

assuming relative magnetic permeability (µr) is near unity.Attenuation of electromagnetic signals is governed by σ and εr. In low-loss material, the

original pulse amplitude (A0) decreases exponentially with depth (z) according to A = A0 e−αz,where the attenuation constant α = 0.5σ

√µ/ε (Theimer et al., 1994). This relationship is valid

only for homogeneous sediments without impedance contrasts. When a propagating electro-magnetic wave encounters a discontinuity in electric, magnetic or conductive properties, part ofthe electromagnetic energy is reflected, the reflection strength being proportional to the magni-tude of change. For a perpendicular incident wave, coming from a medium characterized by awave impedance Z1 and going to a medium with impedance Z2, the reflection coefficient (RC)is expressed as (Brewster and Annan, 1994):

RC =Z2 −Z1

Z2 +Z1. (2.5)

With the same parameters present in the calculations of electromagnetic wave impedance (Equa-tion (2.1)) and electromagnetic wave velocity (Equation (2.2)), velocity contrasts can be takenas a measure for impedance contrasts. If one assumes σ insignificant and contrasts in µ negligi-ble, Equation (2.5) can be rewritten as:

RC =√

εr2 −√εr1√

εr2 +√

εr1. (2.6)

2.3 Methods

2.3.1 Ground-penetrating radar

In this study, a Sensors&Software pulseEKKO 100 GPR system, consisting of a transmittingand receiving antenna connected to a console and laptop computer (Figure 2.1a) was used. Thetransmitted electromagnetic pulse is ideally meant to penetrate the subsurface in a beam asnarrow as possible. Some of the energy, however, travels directly to the receiving antenna asairwave and groundwave, giving the signal at the top of the resulting radar section (Figure 2.1b).

14


Z = x

Z = y

Data storageSystem control

Transmittingantenna

Receivingantenna

Layer I

Layer II

Dep

th

Horizontal position

Tim

e [n

s]

(b)(a)

Figure 2.1: (a) Ground-penetrating radar setup and method and (b) resulting radar section.

Part of the remaining energy, which enters the subsurface, reflects at layers of changing elec-tromagnetic wave impedance and travels back to the receiver. The quantity of energy receivedand the associated arrival time are stored in the computer. The lateral extent and morphology ofreflectors can be delineated by moving the portable equipment across the surface. The resultingradar section, on which each measurement point is represented by a trace, shows time alongits vertical axis and position along its horizontal axis. The velocity of radar waves in differ-ent layers can be calculated through common-mid-point (CMP) measurements (Arcone, 1984),allowing conversion of travel time to actual depth.

The electrical conductivity of a material influences penetration depth as well as resolution.Low-conductivity materials, such as unsaturated and coarse-grained sediments, cause little at-tenuation and under ideal circumstances, penetration is on the order of tens of meters (Davis andAnnan, 1989). However, wave velocity and length are highest in low-conductivity materials,leading to a decrease in resolution (Table 2.1). Penetration depth and resolution are also influ-enced by the GPR frequency used for measurement. Lower antenna frequencies are favorablefor penetration, but result in a decrease in resolution. Resolution is commonly assumed to be a

Table 2.1: GPR setup and wave parameters. The wavelengths in dry, moist, and saturated sand are basedon an electromagnetic wave velocity of 0.15 m ns−1, 0.12 m ns−1, and 0.06 m ns−1, respectively. Theresolution was estimated using the generally accepted one-quarter wave length axiom.

System f Antenna Dry sand Moist sand Wet sand[MHz] separation λ Resolution λ Resolution λ Resolution

[m] [m] [m] [m] [m] [m] [m]

PE 100 100 1.00 1.500 0.375 1.200 0.300 0.600 0.150PE 100 200 0.50 0.750 0.188 0.600 0.150 0.300 0.075PE 1000 450 0.25 0.333 0.083 0.267 0.067 0.133 0.033PE 1000 900 0.17 0.167 0.042 0.133 0.033 0.067 0.017f = frequency; λ = wave length; PE = pulseEKKO.

15

Chapter 2

Data storage

(a)

System controlTektronix cable tester

Epoxycasing

Steelrods

Lp

Time [ns]

Ref

lect

ion

coef

ficie

nt

Δts

(c)(b)

Figure 2.2: (a) Diagram of time-domain reflectometry (TDR) equipment, (b) field example, and (c)resulting wave form. The travel time (Δts) along the probe, with length Lp, determines electromagneticproperties (Equation (2.7)).

quarter of the GPR wave length, and ranges from about 2 cm for saturated sands and 900-MHzantennas to almost 40 cm for dry sands and 100-MHz antennas (Table 2.1).

2.3.2 Time-domain reflectometry

The TDR method is based on the propagation velocity of an electromagnetic signal along asediment probe, and was developed to characterize the water content of soils using the relativepermittivity (Topp et al., 1980). Assuming variations in σ are negligible, the propagation ve-locity can be calculated from Equation (2.3). The product of the relative permittivity and therelative magnetic permeability is calculated from the travel time (Δts) of the TDR signal and thelength (Lp) of the probe (Roth et al., 1990):

εrµr =(

c0Δts2Lp

)2

. (2.7)

Factors other than water content, such as soil density, texture and temperature, have a negligibleinfluence on the relative permittivity (Topp et al., 1980). Using these assumptions and µr = 1,the volumetric water content (θ) is found by substitution of εr in the empirical relationship(Topp et al., 1980):

θ = −5.310−2 +2.9210−2εr −5.510−4ε2r +4.310−6ε3

r . (2.8)

Early laboratory studies (e.g., Fellner-Feldegg, 1969) used sediment columns in coaxial trans-mission lines, whilst field probes were later developed and improved (Dalton et al., 1984;

16


Zegelin et al., 1989; Brisco et al., 1992; Heimovaara, 1993). In this study, the TDR equip-ment developed by Heimovaara and Bouten (1990) was used (Figure 2.2a). A 0.05-m-long,three-rod probe, connected to a Tektronix cable tester and laptop computer for system controland data storage, was pushed into the sediment. The cable tester transmits a fast-rise voltagepulse through the transmission line and probe. The frequency band ranges from 300 kHz to3 GHz (Heimovaara et al., 1996), which encompasses all GPR frequencies. At changes in elec-tromagnetic properties, part of the pulse is reflected to the cable tester. After calibration for theepoxy casing, the two reflection points of the resulting wave form (Figure 2.2c) give the traveltime along the rods (Δts), necessary to obtain εrµr (Equation (2.7)).

2.3.3 Sedimentological and textural characteristics

Sedimentary structures were studied both directly in the field and by using lacquer peels andthin sections. The lacquer peels were made to study sedimentary structures and parametersmacroscopically. On these lacquer peels, sedimentary structures or bedding planes appear asnarrow ridges or lows, owing to different grain and pore sizes. Other elements, like organicmaterial and iron oxide, appear as (root) relics or as color changes on the lacquer peel. Thinsections, made from undisturbed sediment samples, were used to study sediment propertiesmicroscopically. Two samples were taken for grain-size analyses.

2.3.4 Synthetic radar traces

Using TDR and sedimentological data, impedance models of the subsurface can be constructed.The models, in which values for depth and εrµr are given, form the input for pulseEKKO soft-ware (Sensors&Software, 1996) that constructs synthetic GPR traces. The program assumesvertically incident electromagnetic waves and calculates all generated reflections, includingmultiples and interlayer reflections. The program does not account for frequency dependenceof electromagnetic properties. Therefore, any possible wave dispersion as a result of variationsin velocity and attenuation within the transmitted frequency bandwidth is not modeled. Firstly,the program transforms the impedance model from a depth scale into a time scale. Then, theimpulse response for the layered model is computed and a correction for spherical-waveformspreading losses is applied. The ground response is obtained by convolution of a standardpulseEKKO wavelet (both 100 and 200 MHz) with the impulse response. Attenuation differ-ences and dispersion were not incorporated into the model. To ensure that small reflections werealso visualized, the 1-D synthetic traces were plotted with an automatic gain control (AGC), ap-plying a gain inversely proportional to the signal strength. A gain limit was applied to preventvery small signals from producing very large gains.

2.4 Sedimentology and stratigraphy of the study site

Fieldwork was conducted in the Boudewijn quarry in Ossendrecht (Figure 2.3). This 15-m-deepquarry can be subdivided into two units; a lower unit of tidal deposits and an upper unit of eoliandeposits.

The lower unit has a thickness of approximately 7.5 m and consists of tidal sediments (Tege-len Formation) deposited during several Tiglian interglacials of Early Pleistocene age (Kasse,1988; Figure 2.4). An unconformity, possibly formed during the regression associated with the

17

Chapter 2

Amsterdam

ArnhemRotterdam

Antwerp

Study site

RhineMeuse

Scheldt

0 25km

N

Figure 2.3: Map of The Netherlands showing location of study area.

Beerse Glacial, separates two subunits. The lower subunit (Hoogerheide Member) consists oflarge-scale cross-bedded very fine to fine sand. Locally, numerous small ripple marks with claydrapes, resulting from frequent flow reversals, are present (Kasse, 1988). The upper subunit(Woensdrecht Member) consists of large-scale cross-bedded very fine to medium sand. The topof this subunit is a laterally continuous clay layer with a maximum thickness of 3 m (Kasse,1988). At the end of the Tiglian interglacial, the climate became cooler and sea level fell. Asa result, a substantial hiatus, locally marked by a coarse-grained residual deposit, is present ontop of the Woensdrecht member at the base of the upper eolian unit. This hiatus represents anextended time of erosion and non-deposition (Kasse, 1988).

The upper quarry unit has a maximum thickness of 7.5 m and consists of eolian sands of thePleistocene Twente and Holocene Kootwijk Formations (Kasse, 1988). These Formations areseparated by a peat layer of the Griendtsveen Formation in blowout hollows or by a paleosolof Podsol type in adjacent higher areas (Figure 2.4). The sand from the Twente Formationdates from the Late Dryas Stadial to the Early Holocene (Schwan, 1991). Locally, a thin soilor peat layer from the Allerød Interstadial is present near the bottom of this subunit. The sandfrom the Twente Formation originates from the banks of the river Scheldt, which flowed justto the southwest of the study area (Figure 2.3). This sand was deposited as sand sheets anddunes and has a large- to medium-scale cross-bedded character. The prominent peat bed andpaleosol separating the Twente and Kootwijk Formations represents most of the Early- andMiddle Holocene, a timespan of approximately 6000 years (Schwan, 1991). Deposition of thesand from the Late-Holocene Kootwijk Formation started as early as 3000 years ago throughbuildup of small dunes (Schwan, 1991). In places, lamination has been obliterated by abundantroots that can be assigned to several discontinuous soil levels.

18


10

15

20

clay

silt

150

210

300

μm

105

Lithology &sedimentarystructures

vv

vv

v

v

v

v vv v

3.000+/-30

9.050+/-45

11.240+/-50

Key:

Large- & medium-scalecross-stratification

Low-anglecross-stratification

Parallel lamination

Lenticular bedding

Small-scalecross-stratification

(Wavy) sand-claybedding

vvv

vPeat

Soil

Roots

Coarse-grainederosional residue

Unconformity

Ad

Late

Dry

as

Wei

chse

lian

Late

Gla

cial

Sta

geLa

te H

oloc

ene

EM

HT

iglia

nC

5T

iglia

nC

3

Epo

chH

OLO

CE

NE

PLE

IST

OC

EN

E

Lith

o-st

ratig

raph

yK

ootw

ijk F

orm

atio

nG

FT

wen

te F

orm

atio

nW

oens

drec

ht M

embe

r

Teg

elen

For

mat

ion

Hoo

gerh

eide

Mem

ber

Ele

vatio

n [m

] abo

ve N

AP

(D

utch

Ord

inan

ce D

atum

)

Age

[yea

rsB

P]

1 4 C sample

Figure 2.4: Schematic stratigraphic column for quarry Boudewijn, Ossendrecht. Data from Kasse(1988), Schwan (1991), and the present study. GF = Griendtsveen Formation, EMH = Early and MiddleHolocene, Ad = Allerød.

The steplike excavation of the quarry allowed investigation of all units, both from the topand the side wall. In this study, a section in the eolian unit was chosen for detailed analysis(Figure 2.5) because: (1) it is made up of highly resistive sand; (2) the paleosol between the twoeolian subunits, and the transition to the Tegelen clay at the base of the eolian sand, were ex-pected to give clear GPR reflections and provide good reference points in the radar sections; (3)the variety of sedimentary structures was considered important in establishing which featurescause reflection of GPR waves; and (4) the terrain at this site was easily accessible and even.

19

Chapter 2

Kootwijk Formation

Twente Formation

Tegelen Formation

Lacquer peel OSL02TDR-section 2



x = 12.5 mx = 4.5 m

(a)


True dip 140o/13o

Sand-sheetfacies A

Twente Formation

Tegelen Formation

Kootwijk FormationLacquer peel OSL01TDR-section 1


4.2 m

1.8 m

Sand-sheetfacies B

Boundingsurface

Dune cross-stratification

12o

26oTrue dip 350o/37o

Paleosol B2Paleosol B1

Anthropogenic sediment

x = 4.5 mx = 12.5 m

120oN

(b)

Figure 2.5: (a) Picture of section wall showing the locations of lacquer peels and TDR sections.(b) Schematic representation of sedimentary and pedogenic structures. In the Twente Formation, leftof lacquer peel OSL02, several diagenetic iron-oxide bands that cross-cut depositional bedding are prob-lematic for the interpretation of sedimentary structures.

20


The schematic representation of sedimentary and pedogenic structures in the studied sec-tion (Figure 2.5b), reflects the general stratification for the upper quarry unit given above. Inthe section studied, the Allerød soil is absent. Three inorganic facies, deposited on a gradu-ally drying depositional surface (Schwan, 1991), comprise the Twente Formation: (1) a basalsub-horizontally bedded sand-sheet facies, (2) a gently dipping sand-sheet facies and (3) a cross-stratified dune facies. Sand-sheet facies B, at the base of the Twente Formation, is character-ized by alternation of medium sand and finer-grained beds. This facies reflects a varying windregime, a damp depositional surface (caused by the underlying semi-pervious Tegelen Forma-tion) and the availability of both sand and silt in the source area. Sand-sheet facies A, locatedabove the basal facies, is characterized by meter-scale gently dipping sigmoidal strata, havinga maximum dip angle of 140◦/13◦. The dune facies at the top of the Twente Formation is sep-arated from sand-sheet facies A by a coarse-grained bounding surface. The dune facies stratahave a maximum dip angle of 350◦/37◦. However, the strike of the strata show considerablevariability, in line with the inferred parabolic shape of the dunes. The top of this cross-stratifiedsand is marked by a 0.3-m-thick paleopodsol (Paleosol B1, Figure 2.5b; hidden by a small ledgeon Figure 2.5a). Except for some faint horizontal lamination, sedimentary structures in the over-lying sand from the Kootwijk Formation are obliterated by several soil levels (e.g., Paleosol B2,Figure 2.5b).

2.5 Results


GPR data were sampled along a line that was located 5 m from the section wall (Figure 2.6) andhad a length of 35 m. The line was sampled with two frequencies (Table 2.2). On the 100-MHzradar section (Figure 2.7a), some of the largest features of the unit are visible. At a depth ofabout 6 m, two gently left-dipping reflections represent the boundary between the eolian sandsand the tidal clays. These reflections become discontinuous and less pronounced towards theleft, which may be a result of increasing reflector depth or a decreasing change in grain-sizeor water content at this reflector. Below the boundary, the electromagnetic signal is attenu-ated quickly and hence no deeper reflectors can be seen (©1 , Figure 2.7a). The sub-horizontalreflections above the sand-clay transition (©2 , Figure 2.7a) originate in sand-sheet facies B (Fig-ure 2.5b). Above this GPR facies, numerous discontinuous and mostly right-dipping reflections(©3 , Figure 2.7a) mark sand-sheet facies A (Figure 2.5b). The gently left-dipping reflections(©4 , Figure 2.7a) that overly this unit mark the cross-stratified dune facies (Figure 2.5b). Thepaleosol between the Twente and Kootwijk Formations (Figure 2.5b) is delineated by the con-tinuous, dipping reflection at an approximate depth of 1.5 m (©5 , Figure 2.7a). At the top of

Table 2.2: Characteristics and locations of GPR measurements.

Line Distance f Mode X0 Xmax Xtot ΔXto wall [m] [MHz] [m] [m] [m] [m]

OSR01 5 100, 200 Reflection 0 35 35 0.25100, 200 CMP 11 - 10, 7.8 0.20

f = frequency.

21

Chapter 2

0

4.51112.535

Radar line OSR01

Section wall

Common-mid-pointmeasurement

distance [m]

50 10

N

Lacquer peel OS

L01Lacquer peel O

SL02

Lacquer peel OS

L03

Figure 2.6: Sketch to illustrate location of measurements in the Boudewijn quarry.

the radar section, the two horizontal reflections (©6 , Figure 2.7a) represent the direct air- andgroundwaves.

On the 200-MHz radar section (Figure 2.7b), more detail is visible, especially in the 2 to4-m depth range. However, depth of penetration has decreased and both the sand-clay transi-tion and sand-sheet facies B are only barely visible in the profile (©1 and ©2 , Figure 2.7b). Thereflections representing sand-sheet facies A (©3 , Figure 2.7b) generally dip to the right as ex-pected (Figure 2.5b). Left-dipping reflections (©4 , Figure 2.7b) mark the overlying cross-beddeddune facies (Figure 2.5b). The paleosol at the base of the Kootwijk Formation (Figure 2.5b) isrepresented by the reflection at a depth of around 1.5 m (©5 , Figure 2.7b). The horizontal re-flection between the paleosol and the two direct waves (©6 , Figure 2.7b) may represent either(horizontal) bedding or one of the discontinuous soil horizons B2 (Figure 2.5b).

2.5.2 Textural characteristics

Laser grain-size analyses (Konert and Vandenberghe, 1997) and thin sections (from lacquer peelOSL01) provide information on the sediment characteristics. The sandy intervals consist of fineto medium sand (229.1 µm to 287.5 µm) that is subrounded and moderately to well sorted.Pore space occasionally exceeds maximum grain-size, indicating rooting or burrowing. In thepaleosol a large amount of organic material (∼5% of total sediment volume) occupies part ofthe pore space.

Table 2.3: Locations and sizes of lacquer peels.

Name Ytop [m] Ybot [m] Ytot [m] X [m]

OSL01 0.0 2.3 2.3 12.5OSL02 2.1 4.7 2.6 11.0OSL03 3.7 5.6 1.9 4.5

22


Gain: AGC 100. Filters: 3 down the trace; 2 trace to trace

0

2

4

6

8

0

2

4

6

8

Position [m]20 18 16 14 10 8 6 4 2 012

TegelenFormation

KootwijkFormation

TwenteFormation

14 12 10 8 6 4Position [m]

0

2

4

6

8

1

3

6

7

0

2

4

6

8

0

2

4

6

8

Position [m]20 18 16 14 12 10 8 6 4 2 0

Gain: AGC 100. Filters: 4 down the trace; 2 trace to trace

TegelenFormation

KootwijkFormation

TwenteFormation

14 12 10 8 6 4Position [m]

0

2

4

6

81

56

7

(b)

(a)

4 4

3

4 45

3

2

3

2

2

200 MHz

100 MHz

Dep

th [m

]

Dep

th [m

]D

epth

[m]

Figure 2.7: Radar sections for line OSR01 (left) and interpretation (right); (a) 100 MHz, (b) 200 MHz.The outlined part corresponds with the interpreted section shown in Figure 2.5. CMPs indicate thatthe electromagnetic wave velocity was 0.11 m ns−1 in the upper highly resistive subunit, and averaged0.09 m ns−1 for the total unit. Radar sections for line OSR01 are printed with a velocity of 0.09 m ns−1.Since this velocity is an average for the entire unit, the plot is stretched and compressed in its higher andlower parts respectively. Average vertical exaggeration is 1.4. The numbered labels in the interpretedradar sections refer to the text and represent: ©1 , tidal clays, ©2 , sand-sheet facies B, ©3 , sand-sheet faciesA, ©4 , dune facies, ©5 , paleosol, ©6 , overlapping air- and groundwaves, and ©7 , reflections dipping in theopposite direction to the sedimentary structures.

2.5.3 Time-domain reflectometry and lacquer peels

Lacquer peels were taken from representative parts of the wall covering the full height of thesection (Table 2.3) and TDR data (Table 2.4) and sediment samples (Table 2.5) were then col-lected from these transects (Figure 2.5). Comparison of TDR diagrams with associated lacquerpeels (Figure 2.8) illustrates several relationships between sediment characteristics and elec-tromagnetic properties. The product εrµr (Equation (2.3)) increases from a baseline value ofabout 4 in the upper part of the section to almost 25 in its lower part. The baseline value of4 represents the relative permittivity of dry sand, whereas the value of 25 is characteristic ofthe relative permittivity of water-saturated sand (Davis and Annan, 1989; Wensink, 1993), bothhaving a relative magnetic permeability of unity. Following Equation (2.4), these values for εr

23

Chapter 2

correspond to volumetric water contents of 0.055 and 0.4, respectively. Numerous excursions inεrµr are superimposed on the general trend and most of these can be linked to features recordedin the lacquer peels.

The paleosol at a depth of 1.8 m gives an increase in εrµr to about 8 (Figure 2.8a). This iscaused by the fact that organic material, present in this paleosol, holds water, which increasesrelative permittivity. Figure 2.8b shows a similar response to organic matter when two indi-vidual cross-sets, with windblown organic material, cause the relative permittivity to increase.In the lower part of Figure 2.8b and the upper part of Figure 2.8c, several iron-oxide bands,identified in the lacquer peels, result in sharp increases in εrµr. Most of these iron bands followdepositional bedding, but some, such as the band at a depth of 4.4 m (Figure 2.8b), cross-cut thebedding, indicating a diagenetic origin of the iron bands. Since the εrµr peaks typically occurdirectly on top of these cemented and less permeable layers, they may be caused by stagnatingwater occupying pore space, resulting in an increase of εr. Alternatively, an increase in therelative magnetic permeability, caused by the presence of iron-oxide minerals, may bring aboutthe observed peaks in εrµr. The small iron-oxide-related excursions in the TDR log associatedwith lacquer peel OSL02 (Figure 2.8b) are superimposed on a baseline value for εrµr of about 4(volumetric water content of 0.055). In the TDR log associated with lacquer peel OSL03 (Fig-ure 2.8c), the amplitude of the iron-oxide-related excursions is higher. This suggests that, sincewater saturation increases downward in lacquer peel OSL03, stagnating water and thus relativepermittivity is the main cause for the increase in εrµr.

In the zone where average water content remains low and changes in εrµr are caused byeither organic material or iron oxide (Figures 2.8a and 2.8b), it is noteworthy that none of thesechanges can be related directly to grain-size. In lacquer peel OSL02, large-scale cross-beddingis obvious but none of the bedding planes cause εrµr to change significantly, except whereorganic material or iron is present. For example, the gravel layer at 2.85-m depth does not resultin any change in εrµr in the TDR log (Figure 2.8b). In contrast, εrµr fluctuates significantlywhere water saturation increases (Figure 2.8c). In this case, fluctuations that coincide with thepresence of iron-oxide bands as well as sedimentary structures are superimposed on a generalincrease in εrµr with depth. Here, fine-grained layers from sand-sheet facies B (Figure 2.5)enhance capillary forces and therefore support changes in relative permittivity. Thus, GPRreflections from sedimentary structures, not related to mineralogical change, can be expectedonly when values for relative permittivity and volumetric water content exceed the baselinevalues of 4 and 0.055, respectively.

Table 2.4: Characteristics and locations of TDR measurements (see Figure 2.5).

Name TDR section Y0 Ymax Ytot ΔY Lp X Location in grid[m] [m] [m] [m] [m] [m]

z970605.001 1 0.8 1.95 1.15 0.050 0.10 12.5 Lacquer peel OSL01z970605.002 1 0.8 2.00 1.20 0.050 0.05 12.5 Lacquer peel OSL01z970606.003 1 1.6 2.30 0.70 0.025 0.05 12.5 Lacquer peel OSL01z970606.004 2 2.1 4.70 2.60 0.050 0.05 11.0 Lacquer peel OSL02z970606.005 3 3.7 6.00 2.30 0.050 0.05 4.5 Lacquer peel OSL03Lp = TDR probe length.

24


2.5.4 Synthetic radar traces

Using the three TDR logs (Figure 2.8), one log covering the full section was compiled. Fromthis composite TDR log and the lacquer peels, giving information on εrµr and sediment char-acteristics respectively, four impedance models were constructed, each focusing on a particularcharacteristic of the section (Figure 2.9). Using the four impedance models, 1-D synthetic GPRtraces were constructed. The layered impedance models have a depth scale whereas the syn-thetic traces have a time scale. Since the product of relative permittivity and relative magneticpermeability, and thus the wave velocity, varies over depth, the relative thickness of the layersalso varies.

Figure 2.9a shows the graph of εrµr versus depth that was used as input for the differentimpedance models. Figure 2.9b represents the simplest situation with a basic value of 4 and twoexcursions: the paleosol and the increase in εrµr above the sand-clay transition. In the syntheticradar traces, all layer boundaries appear as radar reflections, and there are also multiples ofearlier reflections (indicated by arrows). The situation in Figure 2.9c represents the slightlymore complex paleosol (as found in the TDR logs) and a stepped increase in εrµr towardsthe sand-clay transition. The thin layer, with a εrµr value of 4, within the paleosol leads tooverlapping waveforms (especially in the 100-MHz synthetic trace). Figure 2.9d focuses onthe small excursions in the TDR logs caused by the iron-oxide bands and windblown organicmatter. In the 200-MHz synthetic trace, all events can be distinguished, but in the 100-MHzsynthetic trace, overlapping waveforms produce a more complicated pattern. The sedimentmodel combining Figures 2.9c and 2.9d leads to highly complicated synthetic radar traces withoverlapping waveforms and multiples (Figure 2.9e).

2.6 Discussion and conclusions

Various reasons exist for changes in εrµr and thus for radar reflections, but changes in relativepermittivity are the most important factor. The relative permittivity is predominantly controlledby water content, which is due to the large εr contrast between air and water. Relative per-mittivity is 1 for air and 80 for water, whereas it is 4-6 for most sedimentary minerals (Davisand Annan, 1989). The volume fractions of the different constituents that form the sedimentdetermine εr. Consequently, the sediment porosity as well as the ability of the sediment to holdwater are important factors controlling the relative permittivity (Knoll and Knight, 1994). Whenwater content is high, such as in the lower part of the section (Figure 2.8c), smaller pores causeboth water content and relative permittivity to increase. When water content is low, such asin the higher part of the section (Figures 2.8a and 2.8b), the relative permittivity is controlledby the ability of the sediment to retain water. Both organic material, which holds water, andfine-grained sediment, which enhances capillary forces, cause water content as well as rela-

Table 2.5: Locations of thin section samples.

Name Y [m] X [m] Location in grid

OSS1 0.6 12.5 Lacquer peel OSL01OSS2 1.6 12.5 Lacquer peel OSL01OSS3 1.6 12.5 Lacquer peel OSL01

25

Chapter 2

ε rμr

Soi

l

(a)

05

100 1 2

0.5

1.5

ε rμr

Iron

bandIron

band

Iron

band

Hum

icm

ater

ial

G

rave

l

(b)

3 4

05

10

2.5

3.5

4.5

Depth [m]

Depth [m]

Depth [m]

ε rμr

Iron

ban

d

Iron

ban

d

Incr

ease

inw

ater

con

tent

(c) 15

2025

4 5

05

10

4.5

5.5

Figure 2.8: Lacquer peels and associated TDR measurements. (a) TDR section 1 and lacquer peelOSL01, (b) TDR section 2 and lacquer peel OSL02, and (c) TDR section 3 and lacquer peel OSL03.Lacquer peels are printed in mirror image, such that the structures are shown in their true orientation inthe quarry wall.

26


1 2 3 4 5 6010

20

(a)

(d)

(b)

(c)

(e)

4 6 4 185

56

4 6 4 18

100

200

0

4 64

6

4 8 15 21

4 64 6

4 8 15 21

5

56

100

200

0

100

200

0

100

200

0

200

MH

z10

0M

Hz

200

MH

z10

0M

Hz

200

MH

z10

0M

Hz

200

MH

z10

0M

Hz

Time [ns]

Time [ns]

Time [ns]

Time [ns]

Depth [m]

ε rμr

Fig

ure

2.9:

(a)

Com

posi

teT

DR

log

for

the

com

plet

equ

arry

sect

ion.

Lay

ered

impe

danc

em

odel

s,co

nstr

ucte

dfr

omth

islo

g,an

dre

sulti

ng1-

Dsy

nthe

ticra

dar

trac

es;

(b)

sim

ple,

(c)

soil,

(d)

iron

and

win

dblo

wn

orga

nic

mat

ter,

and

(e)

com

bina

tion.

Arr

ows

indi

cate

mul

tiple

s.N

ote

that

vert

ical

scal

esar

ein

dept

hfo

rth

eT

DR

log

and

laye

red

sedi

men

tmod

els

and

intim

efo

rth

esy

nthe

ticra

dar

trac

es.

27

Chapter 2

tive permittivity to increase. The role of relative magnetic permeability in the GPR reflectionprocess is less clear. Von Hippel (1954) reported a relative magnetic permeability of 1.09 formagnetite-bearing sediment, which would result in an increase in εrµr from 4 to 4.4 for drysandy sediment. When viewing this result in light of the data from the present study in whichiron-oxide-related excursions in εrµr from 4 to about 6 were found (Figure 2.8b), one mustconclude that most of the increase is caused by increasing εr, and thus water content, ratherthan µr. The required relative magnetic permeability values of 1.5 have not been reported inthe literature. This suggests that the possible influence of iron oxides, enhancing µr, is muchsmaller than that of water, enhancing εr.

In the zone with a baseline value of 4 for εrµr, none of the changes can be related directlyto grain size (Figures 2.8a and 2.8b). This explains the poorly defined boundary (i.e., gravellayer in Figure 2.8b) between the cross-stratified dune facies and sand-sheet facies A on theGPR images. Thus, when volumetric water content does not exceed 0.055, grain size is notimportant as a cause of GPR reflections. However, numerous reflections in the zone just belowthe paleosol reflect sedimentary bedding of sand-sheet facies A and the dune facies (Figure 2.7).Both windblown organic material and iron-oxide bands increase εrµr and could cause thesereflections. Some reflections, however, dip in the direction opposite to that of the sedimentarystructures. This is shown for example by the reflection at position 9 m at 4 meters depth inthe 100-MHz section (©7 , Figure 2.7a) and the reflection at position 15 m at 3 meters depth inthe 200-MHz section (©7 , Figure 2.7b). These reflections are most likely caused by iron-oxidebands that cross-cut sedimentary bedding.

The TDR logs could be used to calculate detailed synthetic radar traces. Owing to thenon-vertical quarry wall, the depth in the impedance model, and consequently the time in thesynthetic GPR traces, is overestimated by about 20 percent. Nevertheless, the synthetic traces(Figure 2.9) can be compared with the actual GPR sections (Figure 2.7). Waveforms that overlapin part in the synthetic traces (Figure 2.9b), due to the complex paleosol, are also present in theactual GPR sections. This is well exemplified in the 200-MHz section between position 10 and20 meters at a depth of 1 to 1.5 meters (Figure 2.7b). In Figure 2.9d, the small excursions inthe TDR logs caused by the windblown organic matter and iron-oxide bands were modeled.This zone is difficult to compare with the actual radar sections as the variability in thicknessand intensity of the iron-oxide bands is high and the changes in εrµr are small. Nevertheless,the reflection strengths in the synthetic traces and actual GPR sections match quite well. Theincrease in water content towards the sand-clay transition was modeled in the synthetic tracesusing a simplified three-step increase in εrµr (Figure 2.9c). The three reflection events in thesynthetic traces that result from this tripartition show a clear decrease in amplitude with depthbut cannot be linked to the actual radar sections because the TDR logs show a higher variabilityin εrµr (Figure 2.9a). Nevertheless, synthetic GPR traces can help in understanding the variousreflections in actual GPR sections.

Not all GPR reflections can be matched with events on TDR logs, lacquer peels and 1-Dsynthetic images. The internal lateral variability of the facies is the main problem in correlatingthese data sets. Within the 5-m distance from the GPR sections to the quarry face, both thegeometry and the depth of the facies, as well as the water content, could have changed laterally.In spite of the above, this study shows that the combination of TDR and sedimentologicalanalysis is a very useful technique for qualifying and quantifying electromagnetic properties inthe subsurface and achieving a better insight into the origin of GPR reflections.

28

3

Iron oxides as a cause of GPR reflections

AbstractIron oxides frequently occur as secondary precipitates in both modern andancient sediments and may form bands or irregular patterns. Time-domainreflectometry (TDR) field studies show that goethite iron-oxide precipitatessignificantly lower the electromagnetic wave velocity of sediments. Measuredvariations in magnetic permeability do not explain this decrease. The TDRmeasurements and a dielectric mixing model also show that neither electricalconductivity nor relative permittivity of the solid material are altered signifi-cantly by the iron-oxide material. From drying experiments during all of themeasurements, the amount of iron oxides appears to correlate with the vol-umetric water content, which is the result of differences in water retentioncapacity between goethite and quartz. These variations in water content con-trol relative permittivity and explain the observed variation in electromagneticwave velocity. Using 2-D synthetic radar sections we show that the pattern ofiron-oxide precipitation may have a profound influence on the GPR reflectionconfiguration and can cause major difficulties in interpretation.

This chapter is based on Van Dam, R. L., Schlager, W., Dekkers, M. J., and Huisman, J. A. (accepted). Ironoxides as a cause of GPR reflections. Geophysics, scheduled for publication in the March/April 2002 issue, 67(2).Reprinted with permission.

Chapter 3

3.1 Introduction

Over the last decade, the interest in ground-penetrating radar (GPR) as a tool to investigate theshallow subsurface has widened. During this period, theoretical research focused mainly onsynthetic modeling of wave propagation (e.g., Carcione, 1998). Studies combining wave theoryand geological or petrophysical data have been sparse, and there is a lack of understandingof several features in field data, such as the impact of iron oxides on GPR waves. Recently,layers rich in iron oxides were identified as possible GPR reflectors by Van Dam and Schlager(2000). They used time-domain reflectometry (TDR) and sedimentological analyses to showin a qualitative way that iron oxides significantly lower the electromagnetic wave velocity ofsediments. The purpose of this paper is to quantify this relationship in order to understand whyand how iron-oxide precipitates influence the electromagnetic field, and to establish the possibleimplications for GPR records.

Several field and laboratory studies show that iron oxides can have a profound influence onelectromagnetic wavefields in general (Robinson et al., 1994; Klein and Santamarina, 2000),magnetic susceptibility (Mullins, 1977), or attenuation (Matzler, 1998). Iron-oxide precipitatesare found in various chemical forms in a wide range of depositional and climatic environments(Driessen and Dudal, 1991). Iron is easily soluble under reducing conditions, and easily trans-ported after reduction to Fe2+. Oxidation towards Fe3+ and subsequent precipitation of ferrouscompounds occurs once oxygen levels are sufficiently high. The pattern of precipitation usuallyreflects the hydraulic properties of the sediment, such that a fluctuating groundwater level canlead to ’ferricrete’ horizons cemented by iron oxide (Cornell and Schwertmann, 1996), whereasroots and fault zones (Bense et al., in review) may cause irregular patterns of iron-oxide precip-itation .

In this paper, we try to understand why iron-oxide precipitates lower the electromagneticwave velocity of sediments. We sampled sediment with varying degrees of iron-oxide contentin a quarry in The Netherlands, and analyzed the characteristics of the sediment using grainsize analysis, scanning electron microscopy (SEM), chemical analyses by inductively coupledplasma atomic emission spectrometry (ICP-AES), and thermogravimetric techniques (TGA).We studied the effect of the iron oxides on all components of the electromagnetic field usingmagnetic measurements, TDR measurements, and dielectric mixing models. Finally, we usedthe information from these analyses to construct 2-D synthetic radar sections that illustrate thepossible impact of iron-oxide precipitates on GPR reflections.

3.2 Theory

The GPR technique is based on the propagation and reflection of electromagnetic energy in thesubsurface. The (frequency dependent) properties that control the behavior of electromagneticenergy in a medium are dielectric permittivity (ε), where ε = ε0εr, electrical conductivity (σ),and magnetic permeability (µ), where µ = µ0µr (Von Hippel, 1954). Here, ε0 is the permittivityof vacuum (8.85419 10−12 F m−1), εr is the relative permittivity, µ0 is the magnetic permeabil-ity of vacuum (4π10−7 H m−1), and µr is the relative magnetic permeability. Together, theseproperties define the electromagnetic wave impedance (Z) of the medium (Brewster and Annan,1994):

Z =

√jωµ

σ+ jωε, (3.1)

30


where j =√−1 and ω = 2π f is angular frequency [radians s−1]. We do not consider com-

plex permittivity caused by water dipole relaxation because its frequency is above the GPRbandwidth.

The propagation velocity (v) of electromagnetic waves in a medium is found by:

v =c0√

εrµr1+

√1+tan2 δ2

, (3.2)

where c0 is the electromagnetic wave velocity in vacuum (3 108 m s−1) and tanδ is the losstangent. The loss tangent, which equals σ/ωε (see Equation (2.2)), approaches zero for low-loss material, such as sand. Consequently, the velocity is mainly controlled by εr and µr.

When a propagating electromagnetic wave encounters a discontinuity in electric, magneticor conductive properties, part of the electromagnetic energy is reflected. The reflection strengthis proportional to the magnitude of change in electromagnetic properties. For a perpendicularincident wave the reflection coefficient (RC) is expressed as:

RC =Z2 −Z1

Z2 +Z1, (3.3)

where Z1 and Z2 are the impedances of the layers above and below the discontinuity, respec-tively (Brewster and Annan, 1994). The dielectric permittivity, which is generally controlled byvolumetric water content (θ), is assumed to play a dominant role in the reflection process and inthe propagation of electromagnetic waves (Sutinen, 1992; Huggenberger, 1993). The electricalconductivity (σ) of a material depends on the amount of kinetic energy that is irreversibly con-verted into heat (Van Der Kruk et al., 1999) and controls, together with εr, the attenuation of anelectromagnetic wave (e.g., Daniels et al., 1988; Brewster and Annan, 1994; Carcione, 1998).Relative magnetic permeability, which is the third factor influencing the impedance (Equation(3.1)), is generally assumed to be near unity (e.g., Daniels et al., 1988; Brewster and Annan,1994; Carcione, 1998). A range of typical electromagnetic properties for common geologicmaterials was given by Davis and Annan (1989).

3.3 Methods


TDR is widely used to measure the water content of soils. The TDR method uses empiricalrelationships between the relative permittivity (εr) and volumetric water content (θ), under theassumption that µr is equal to 1 (e.g., Topp et al., 1980). The method is based on the propaga-tion velocity of a guided electromagnetic signal in sediment (e.g., Brisco et al., 1992). For themeasurements we used a three-rod probe with a length of 5 cm that was inserted into the sedi-ment. The probe is connected to a Tektronix cable tester that transmits a fast-rise voltage pulse(bandwidth about 300 kHz to 3 GHz) through the transmission line and probe. The waveformresulting from the reflection at the end of the probe is analyzed to obtain the travel time (Δts) ofthe pulse along the rods. This travel time is used to calculate the velocity of the electromagneticwave or can be used to obtain εrµr from the formula (Roth et al., 1990):

εrµr =(

c0Δts2Lp

)2

, (3.4)

31

Chapter 3

where Lp is the length of the probe [m]. In this study, we use the quantity εrµr instead ofvelocity to present and discuss the TDR results. This facilitates comparison with εr values thatare commonly used in TDR and GPR literature.

Dielectric mixing models are frequently used to analyze the contribution of the differentsediment constituents (solid, liquid and gas phase) to εr (e.g., Tinga et al., 1973). Under theassumption that µr is equal to 1, a simple three-phase model can be used for humid samples(Bohl and Roth, 1994):

εr = (θεαmw +(1−η)εαm

s +(η−θ)εαma )1/αm , (3.5)

which reduces to a two-phase model for dry samples:

εr = ((1−η)εαms +ηεαm

a )1/αm . (3.6)

Here η is porosity, εw is relative permittivity of water (80), εs is relative permittivity of solidmaterial, εa is relative permittivity of air (1), and αm is a constant (0.5 for isotropic and homo-geneous material; Roth et al., 1990).

TDR is also used to calculate the bulk electrical conductivity (σDC) of sediment (Giese andTieman, 1975). For this, the attenuation of the voltage pulse is estimated from the reflectioncoefficient at long times (ρ∞) using Rtot = Zc(1 + ρ∞)/(1−ρ∞), where Rtot is the total resis-tance of sediment and equipment [Ω] and Zc is the cable impedance [Ω]. The bulk electricalconductivity can be calculated according to Huisman and Bouten (2000):

σDC =Kp

Rtot − (LcRc +R0), (3.7)

where Kp is a probe constant, Rc is the resistance of the coaxial cable [Ω m−1], R0 is the re-sistance of cable tester and connectors [Ω], and Lc is the cable length [m]. The calculated DCelectrical conductivity is frequency independent but can be recalculated into a real conductivityusing (Ulriksen, 1982):

σ = σDC +ωε′′, (3.8)

where ε′′ is the imaginary part of the permittivity.

3.3.2 Thermogravimetric analysis

Thermal analysis refers to the study of the behavior of materials as a function of temperaturechange. Thermogravimetric analysis is a technique that measures the change in weight of asample as a function of temperature. The shape of the thermogravimetric (TG) curve givesinformation on sample composition, thermal characteristics of the sample and on the productsformed during heating (Blazek, 1973). Parts of the curve with zero slope indicate sample sta-bility for a certain temperature range. Non-zero slopes give the rate and direction of the weightchange. The derivative of the thermogravimetric curve (DTG) shows the change in weight withtime as a function of temperature or time.

3.3.3 Magnetic measurements

In the low-intensity field region, the magnetization M is linear with the intensity of the appliedfield H (or B) for all substances. Therefore, the initial or low-field magnetic susceptibility,

32


defined as χ = M/H or χ = Mµ0/B, is a material-specific property. The magnetic susceptibilityis usually expressed per unit volume (κ) and referred to as mass-specific susceptibility (χ) whenexpressed per unit mass. The applied field, initial susceptibility, and magnetic permeability arerelated as follows:

µ =BH

= µ0(1+κ), (3.9)

andµr =

µµ0

= 1+κ. (3.10)

For the mixed sediments we consider, magnetic interaction is not important and the magneticanisotropy (directional dependence of the initial susceptibility) is estimated to be <1%, so themeasured χ can be seen as the sum of the χ’s of individual minerals.

We used a bridge-type susceptometer (Jelinek, J., 1980, AGICO, Czech Republic), wherethe susceptibility of a sample is related to the imbalance between two coils in one of whichthe sample is inserted. The field strength, H, applied by the instrument is 300 A m−1 at anoperating frequency of 920 Hz. Our samples contain only iron oxides in small amounts, and weneglect any possible frequency dependence (Worm et al., 1993). From the instrument magnetic-susceptibility reading (Rκ) the mass-specific susceptibility χ can be calculated according to:

χ =Rκ10−6kV

m, (3.11)

where k is a range factor (0.05 for the present samples), V is the standard sample volume of10−5 m3, and m is mass [kg]. To convert mass-specific to volume-specific susceptibility values,we used a specific density (ρ) of 2500 kg m−3 in the formula κ = χρ. According to Equation(3.10), κ can now be used to calculate µr, which is an important input parameter for the calcu-lation of the impedance Z (Equation (3.1)) and electromagnetic wave velocity (Equation (3.2)).

3.3.4 GPR modeling

Using TDR data, models of electromagnetic wave velocity in the subsurface can be constructed.These 2-D layer models form the input for pulseEKKO ray-trace modeling software (Sen-sors&Software, 1996) that constructs synthetic GPR sections. The program calculates spatiallycorrect sections but does not include multiples and diffractions to the modeling. Each layer inthe model is assigned a constant velocity and attenuation. We convolved a Ricker wavelet withthe layer-section impulse-response function to form the synthetic traces.

3.4 Sampling and sediment description

3.4.1 Sampling and TDR tests

Data were obtained from moist eolian quartz sand of the Twente Formation in quarry Boude-wijn, Ossendrecht, The Netherlands (Figure 3.1a). We conducted TDR measurements for threeprofiles (Figure 3.1). Site 5a (Figures 3.1b and 3.1c) is made up of an orange zone for the upper0.28 m, a red band, rich in iron oxides, for the next 0.08 m and a white zone at the bottom. TheTDR profile shows a baseline value for εrµr of around 6 in the orange and white sediment. Atthe position of the band rich in iron oxides, the εrµr values increase to around 9, thus causing

33

Chapter 3

Am

ster

dam

050

kmN

Rhine

Sch

eldt

Meuse

Stu

dy

site

Rot

terd

am

(a)

811

5a

2.0m

67

5

2.4m

49

10

NE

SW

Gro

undw

ater

OS

11-1

OS

11-2

OS

11-3

OS

11-4

OS

11-5

0 0.5

1.0

1.5

2.0 2.5

11

610

14

1 2

0.5

1.5

(e)

Depth [m]

ε rμr

0.4

0.8

0.2

0.6

67

OS

8-1

OS

8-2

OS

8-3

0m 1.0m

8(d

)

Depth [m]

ε rμr

68

10

0.2

0.4

0.6

5a

0m

0.7m

OS

5a-1

/2

OS

5a-3

/4

OS

5a-8

/9

OS

5a-1

1/12

OS

5a-6

/7b

OS

5a-6

/7a

(c)

TD

R-p

rofil

e

Ser

ies

of s

ampl

es (

TG

A,I

CP

, mag

netic

s, g

rain

size

)

Bul

k sa

mpl

e (w

hite

san

d)

Bul

k sa

mpl

e (r

ed s

and)

Bul

k sa

mpl

e (o

rang

e sa

nd)

Key

:

Depth [m]

ε rμr

(b)

Figure 3.1: Diagram showing locations of the study area and sampling points and the results of theTDR measurements: (a) location of the study area in The Netherlands; (b) photomosaic of quarry face,showing the position of local groundwater level, bulk-sample locations, and rectangles (not to scale)referring to detailed sampling- and TDR-measurement sites; (c) site 5a; (d) site 8; and (e) site 11.

34


a drop in electromagnetic wave velocity (Equation (3.2)). Site 8 (Figures 3.1b and 3.1d) is lesspronounced in color differences, but has a reddish zone (0.35 - 0.55 m) between orange mate-rial. The TDR profile does not reveal an obvious trend related to the color changes. Site 11(Figures 3.1b and 3.1e) consists of two orange zones (0 - 0.5 m and 1.5 - 2 m), a red zone (0.5 -1.5 m) and a white zone for the lower half meter. The TDR profile for this site shows a clearstep in the εrµr value from around 5 to 6 at the change from orange to red sediment at 0.5 m. Ata depth of 2.0 m, a sharp peak, related to a finer grained layer, is present. Due to the proximityof the water table, the εrµr curve does not return to lower values in the white material.

Using the red-staining of the sediment we identified three units: white, orange, and red. Atan even distribution over the units we took sample material at 14 specific locations and added6 samples from bulk material (Figure 3.1). The samples have an average grain size of 281 µm,with about 1.5% in the silt fraction (Table 3.1). The material is essentially clay free and containsno organic matter. Scanning electron microscopy (SEM) and energy dispersive X-ray spectra(EDX) show that the iron oxide is present as a thin coating around the grains. Using inductivelycoupled plasma atomic emission spectrometry (ICP-AES) we found the amount of iron in thesediments to vary in abundance between 980.2 and 7050.1 PPM (Table 3.1). These valuesare for iron only since (iron) oxides cannot be defined from chemical methods such as ICP-AES. Nevertheless, these values represent relative proportions of the iron oxides in the differentsamples. The red samples form a distinct group with high iron contents, whereas the orangeand white units with lower iron contents show some overlap.

Table 3.1: Description of samples.

Sample code Unit Grain size ICP-AES Thermogravimetric analysis (TGA)(Figure 3.1) Mean Sd Fe content Weight <105 ◦C 200-300 ◦C 400-600 ◦C

[µm] [µm] [PPM] [g] Δm [%] Δm [%]1 Δm [%]1

OS5a-8/9 White 248.8 119.9 1485.8 78.92 3.152 0.029 0.110OS5a-11/12 White 239.3 123.1 1988.9 55.82 1.809 0.042 0.165OS11-5 White 282.6 133.6 980.2 72.01 9.221 0.026 0.084Bulk-whi A White 274.8 135.4 992.9 66.44 3.217 0.028 0.096Bulk-whi B White 271.4 133.7 1296.8 89.76 3.825 0.028 0.114OS5a-1/2 Orange 256.7 126.4 2013.6 64.04 4.265 0.052 0.130OS5a-3/4 Orange 269.6 133.5 2431.6 80.09 3.698 0.044 0.121OS8-1 Orange 298.8 149.2 2162.4 103.30 4.093 0.033 0.099OS8-3 Orange 248.9 132.8 2708.6 46.28 1.239 0.055 0.156OS11-1 Orange 337.2 163.6 1751.9 69.13 4.149 0.035 0.099OS11-4 Orange 307.1 149.5 1951.9 76.04 7.614 0.036 0.090Bulk-ora A Orange 246.6 122.9 2829.9 91.02 5.325 0.056 0.130Bulk-ora B Orange 244.4 122.1 2743.6 84.14 5.516 0.056 0.124OS5a-6/7a Red 283.8 143.3 5762.5 66.68 4.668 0.118 0.136OS5a-6/7b Red 260.2 128.5 5947.0 88.33 10.064 0.122 0.144OS8-2 Red 272.3 138.6 4223.1 60.94 5.069 0.085 0.121OS11-2 Red 360.8 167.2 7050.1 55.16 7.287 0.151 0.112OS11-3 Red 271.2 131.0 5551.0 66.00 7.634 0.105 0.104Bulk-red A Red 325.5 157.7 6450.4 87.75 6.244 0.122 0.107Bulk-red B Red 326.5 157.0 6181.5 82.43 5.756 0.101 0.099Sd = standard deviation; Δm = weight loss.1 corrected for <105 ◦C weight loss

35

Chapter 3

red

orange

white

92

94

96

98

100

0 200 400 600 800 1000

Temperature [ C]o

Wei

ght [

%]

(a)

-0.05

0

0.05

0.1

0.15

0.2

0 200 400 600 800 1000Der

ivat

ive

wei

ght [

%C

]o

-1

Temperature [ C]o

(b)

red

orange

white

Figure 3.2: Results of thermogravimetric analysis. All weights are expressed in percentage of totalweight to facilitate comparison. (a) Thermogravimetric (TG) curves that plot the sample weight againsttemperature. (b) Derivatives (DTG) of the TG curves that show weight change (averaged over 10 mea-surement points) against temperature.

3.4.2 Thermogravimetric analysis

Thermogravimetric analysis was conducted for 20 samples with a typical weight of 55 to 90 mg(Table 3.1) that were selected from the three units discussed before. The samples were heatedfrom about 22 to 943 ◦C at an average rate of 10.2 ◦C min−1. Figure 3.2 shows the resultsper unit as both thermogravimetric (TG) curves and their derivatives (DTG). The TG curvesshow a strong weight loss between the start of the measurements and 100 ◦C. This is causedby the evaporation of free and capillary water from the pores (Blazek, 1973; Foldvari, 1991).The white and the red samples have an average weight loss of about 4 and 7%, respectively,whereas the orange samples occupy an intermediate position close to the white samples. Thesharp peaks in the DTG curves at 100 ◦C are the result of some temperature fluctuation. Anendothermic reaction due to the boiling of water, leading to an instantaneous heat release, causesthis fluctuation.

In order to analyze and to be able to compare the thermal behavior of the samples for thetemperature range above 100 ◦C, we corrected for the weight losses due to water evaporation.The results for these calculations show three weight-loss events (Figure 3.3), which have theirmaxima at about 280, 490, and 570 ◦C, respectively. The losses were calculated in terms ofpercentage for the range 200 to 300 ◦C for the first event and over the range 400 to 600 ◦C for

36


99.6

99.7

99.8

99.9

100

0 200 400 600 800 1000

Temperature [ C]o

Wei

ght [

%]

(a)red

orange

white

0.0000

0.0005

0.0010

0.0015

0.0020

0.0025

0 200 400 600 800 1000

Temperature [ C]o

(b)

Der

ivat

ive

wei

ght [

%C

]o

-1 red

orange

white

Figure 3.3: (a) Thermogravimetric curves (TG) and (b) their derivatives (DTG) showing the results ofthermogravimetric analysis after correction for free-water content, focusing on the thermal behavior ofthe samples for the temperature range above 100 ◦C. Weights have been set to 100% at 105 ◦C to enablecomparison. The DTG curves were calculated as an average over 10 measurement points.

the second (including the third) event (Table 3.1). The derivative curves never reach zero untilabout 800 ◦C. This shows there is a small but continuous background weight loss on whichthe major events are superimposed. We attribute this weight loss to the gradual evaporation ofadhesive water from the grain surface (De Marsily, 1986). The first weight-loss event causes aclear difference in curve shape between the white and the red samples, while the orange samplesoccupy an intermediate position. The white samples lack the presence of this event, whereasthe red samples show a clear weight loss (Figure 3.3). The weight loss in the first event thusseems to depend on the presence of iron oxides in the material. This event shows a very highcorrelation (R2 = 0.97) with the iron contents from the ICP-AES measurements (Table 3.1). Thesecond weight-loss event has the same appearance in all three units and its width and height inthe DTG curves are comparable (Table 3.1). Possibly, this event is related to small amountsof clayey particles present in the sediment. The third weight-loss event at 570 ◦C is small butpresent in all samples. This event is most likely caused by an endothermic reaction, related tothe α-β transformation of quartz (SiO2), which is the main constituent of the sediment (Blazek,1973).

37

Chapter 3

3.5 Results - electromagnetic properties

3.5.1 Magnetic measurements

We measured mass-specific susceptibility (χ) for the 20 samples, both under field conditions(i.e., humid) and after drying at 45 ◦C for 16 hours. The values vary between −0.046 10−8 and0.576 10−8 m3kg−1 for the humid samples, and between 0.001 10−8 and 0.767 10−8 m3kg−1 forthe dry samples (Table 3.2). The lower average susceptibility of the humid samples is causedby the diamagnetic water, having a negative χ. Figure 3.4 summarizes the results for the threeunits in a plot of µr against sample weight. The iron-oxide concentration appears to have a largerinfluence on µr than water does. The offset on the vertical (µr) axis between a white (iron-oxidefree) sample and a red sample (iron-oxide rich) is about three times the offset between a humidand a dry sample of the same unit. The calculated µr values allow a clear distinction among theunits, but the differences in µr are too small to cause a significant contrast in Z (Equation (3.3)).

3.5.2 TDR measurements

In the laboratory, 200-cm3 cylinders and 10-cm TDR probes were used to measure εr and σDC

of the bulk sediment from the three different units (Figure 3.1). We filled three cylinders withhumid sediment (A-samples), and three other cylinders with dry sediment (B-samples). Wemeasured the A-samples both under field conditions and after drying at 70 ◦C for 14 hours.The volumetric water content for the A-samples followed from the weight loss during dry-

Table 3.2: Results of magnetic susceptibility measurements.

Sample code Unit Humid samples Dry samplesWeight χ µr Weight χ µr

[g] [10−8 m3kg−1] [g] [10−8 m3kg−1]

OS5a-8/9 White 9.35 0.06844 1.00000171 8.70 0.15284 1.00000382OS5a-11/12 White 8.99 0.26690 1.00000667 8.44 0.35438 1.00000886OS11-5 White 9.09 −0.03574 0.99999911 8.45 0.03374 1.00000084Bulk-whi A White 8.62 −0.04583 0.99999885 8.18 0.00061 1.00000002Bulk-whi B White 9.62 0.01662 1.00000042 9.08 0.07595 1.00000190OS5a-1/2 Orange 8.51 0.19925 1.00000498 8.04 0.27300 1.00000683OS5a-3/4 Orange 9.30 0.20854 1.00000521 8.71 0.30318 1.00000758OS8-1 Orange 8.83 0.20263 1.00000507 8.12 0.30926 1.00000773OS8-3 Orange 9.28 0.54734 1.00001368 8.67 0.67443 1.00001686OS11-1 Orange 9.55 0.11258 1.00000281 8.85 0.20294 1.00000507OS11-4 Orange 9.40 0.07552 1.00000189 8.42 0.19116 1.00000478Bulk-ora A Orange 10.22 0.28858 1.00000721 9.50 0.37455 1.00000936Bulk-ora B Orange 10.06 0.30916 1.00000773 9.24 0.42628 1.00001066OS5a-6/7a Red 9.96 0.43719 1.00001093 9.08 0.58553 1.00001464OS5a-6/7b Red 8.89 0.57620 1.00001441 7.97 0.76731 1.00001918OS8-2 Red 8.98 0.41602 1.00001040 8.26 0.53691 1.00001342OS11-2 Red 10.33 0.48079 1.00001202 9.48 0.61125 1.00001528OS11-3 Red 9.87 0.38768 1.00000969 8.75 0.56885 1.00001422Bulk-red A Red 11.26 0.48951 1.00001224 10.30 0.63325 1.00001583Bulk-red B Red 10.09 0.53138 1.00001328 9.27 0.66851 1.00001671χ = mass-specific susceptibility; µr = relative magnetic permeability.

38


0.99999

1.00000

1.00001

1.00002

8 9 10 11

Weight [g]

Rel

ativ

e m

agne

tic p

erm

eabi

lity

red wet (N=7)red dry (N=7)orange wet (N=8)orange dry (N=8)white wet (N=5)white dry (N=5)

Figure 3.4: Results of magnetic measurements using a bridge-type susceptometer. The relative magneticpermeability (µr) is plotted against sample weight. The measurements were grouped for white, orangeand red units and were conducted under both field conditions (i.e., humid) and after drying. The errorbars give the standard deviations of the measurements.

ing (Δm) and the volume (V ) of the cylinder (θ = Δm/V ). We calculated the porosity fromη = (1−mdry)/ρV , where mdry is the dry weight and ρ is the density of quartz (2500 kg m−3).The difference in porosity between the A- and B-samples (Table 3.3) reflects a different degreeof packing of the sediment.

The values for σDC (Equation (3.7)) range from 4.31 10−3 to 4.54 10−3 S m−1 for the humidmaterial, and from 1.15 10−4 to 1.91 10−4 S m−1 for the dry material (Table 3.3). The higherconductivities for the humid samples are caused by the water that occupies part of the porespace, replacing the non-conductive air. We observed that σDC showed no trend related to theiron-oxide units. It follows that the amount of iron oxides does not affect the conductivity of thesamples enough to create significant impedance changes between the samples (Equation (3.1)).Similar conductivity values were found for a sediment with 10% iron oxides by weight from

Table 3.3: Results from the laboratory TDR measurements on bulk material.

Sample code Unit Status Textural characteristics Dielectric propertiesWeight η ρb θ σDC εr εs

[g] [m3m−3] [g cm−3] [m3m−3] [S m−1]

Bulk-whi A White Humid 309.28 0.418 1.546 0.092 0.004308 5.119 3.643Bulk-ora A Orange Humid 323.36 0.402 1.617 0.122 0.004537 6.290 3.599Bulk-red A Red Humid 330.56 0.396 1.653 0.144 0.004480 7.616 4.099Bulk-whi A White Dry 290.80 0.418 1.454 - 0.000151 2.309 3.584Bulk-ora A Orange Dry 298.90 0.402 1.495 - 0.000137 2.439 3.763Bulk-red A Red Dry 301.83 0.396 1.509 - 0.000115 2.451 3.751Bulk-whi B White Dry 283.68 0.433 1.418 - 0.000171 2.389 3.848Bulk-ora B Orange Dry 289.50 0.421 1.448 - 0.000191 2.385 3.764Bulk-red B Red Dry 278.11 0.444 1.391 - 0.000178 2.276 3.666η = porosity; ρb = bulk density; θ = volumetric water content; σDC = bulk electrical conductivity;εr = relative permittivity; εs = relative permittivity of solid material.

39

Chapter 3

y = 0.0231x 91563

R 2 = 0.823

0.04

0.08

0.12

0.16

0.99999 1.00000 1.00001 1.00002 1.00003

Relative magnetic permeability (dry)

200-

300

C w

eigh

t los

s [%

]o

whiteorangered

0.00

Figure 3.5: A cross plot of the 200 to 300 ◦C weight loss % from the thermogravimetric measurementsand the relative magnetic permeability (µr). The diagram shows a positive correlation, illustrated by thecorrelation coefficient of the fitted power function, related to the amount of iron oxides in the sediment.

Meerske, The Netherlands (Reckman, J., 2000, private communication). The conductivity ofthis sediment, with a density of 2240 kg m−1, was measured to be around 2 10−4 S m−1.

The calculated εr values for the humid A-samples range from 5.12 to 7.62 and show apositive correlation with the amount of iron oxides in the sediment (Table 3.3). Using dielectricmixing models (Equations (3.5) and (3.6)) the influence of the different phases (solid, water,gas) on εr was estimated. The calculated values for εs, which lie in a narrow range of about 3.6to 3.8, with one peak at 4.1 (Table 3.3), show no correlation with bulk density, water content,and iron-oxide content. The observed spreading of εs values falls within the expected range(Roth et al., 1990) and cannot be interpreted in more detail because we did not measure thesample temperatures, because the density of the solid phase is not known exactly, and becauseof accuracy limitations of the TDR equipment. The calculations using the dielectric mixingmodel show that iron oxides do not directly alter εrµr by changing εs. Instead, the variation inεrµr is caused by changes in volumetric water content, positively correlated with the amount ofiron oxides in the sediment (Table 3.3).

3.6 Discussion

3.6.1 Iron-oxide material

The ICP-AES measurements of iron content and the 200 to 300 ◦C weight loss in the TGAmeasurements show a strong crosscorrelation (R2 = 0.97) and a strong correlation with the 3units (Table 3.1). We assume that both data sets can be used as proxies for the amount of ironoxides in the sediment. A cross plot of relative magnetic permeability (µr) and the 200 to 300 ◦Cweight loss shows a good correlation (Figure 3.5). An exponential trendline was fitted to thedata satisfactorily (R2 = 0.82) on an empirical basis. The plot confirms the observation thatthese variables are controlled by the amount of iron oxides present. Using the information fromboth measurements and data from Blazek (1973) and Mullins (1977) the type of iron oxidecan be determined (Table 3.4). As the samples are composed of several compounds (grains,

40


pore gas and water), the mass-specific susceptibility (χ) is a mixed number. Regarding the verylow χ values measured (Table 3.2), the iron oxides cannot be of ferrimagnetic type such asmagnetite or maghemite (Table 3.4). Antiferromagnets and paramagnets do have the requiredlow susceptibility. Of these, hematite is not a candidate because its DTG peak occurs at atemperature higher than the observed 280 ◦C (Table 3.4). The α- and γ-FeO·OH type ironoxides as well as ferrihydrite, which have the required thermal characteristics, generally occursimultaneously. However, of these goethite is most common and can be expected to be the mainconstituent of the iron-oxide material (Robinson et al., 1994; Cornell and Schwertmann, 1996).

The reaction that occurs for FeO·OH type iron oxide at 280 ◦C is dehydroxylation (Dekkers,1990):

2[FeO ·OH] → Fe2O3 +H2O. (3.12)

This loss of water causes the measured weight loss. Under normal field conditions, the non-polar OH molecules are strongly bound in the crystalline structure and cannot account for thevariation in electromagnetic properties as observed in the Ossendrecht TDR measurements.

3.6.2 Volumetric water content

The magnetic and TDR measurements showed that the goethite has a minor influence on µr, andthat the iron oxides do not significantly change the conductivity or the relative permittivity (εs)of the grains (Table 3.3). The thermogravimetric measurements showed that θ is significantlyhigher for sediments containing iron oxides (Figure 3.2). Figure 3.6, which plots iron-oxidecontent (approximated by the iron content from the ICP-AES measurements) against free-waterrelated weight loss, shows a clear and positive correlation, although the orange and white unitsshow some overlap (Table 3.1). Goethite thus appears to retain more water in the sediment,causing an increase in εr. The observed variations in electromagnetic properties (Figure 3.1)can thus be explained by changes in free-water content. It is important to understand thesedifferences in θ and εr, which can alter the behavior of electromagnetic and GPR waves.

The volumetric water content (θ) for the samples used in the TDR laboratory measurements(Table 3.3) ranges from 0.092 to 0.14. Using the empirical relationship given by Topp et al.(1980), the field measurements of εr (Figure 3.1) give θ values between 0.055 and 0.17. The

Table 3.4: Thermal behavior and magnetic susceptibilities for several iron oxides and soil constituents.Data from Blazek (1973), Mullins (1977), Thompson and Oldfield (1986), Cornell and Schwertmann(1996), and Maher (1998).

Material Chemical Magnetic χ Thermal peakformula status [10−8 m3kg−1] temperature [◦C]

Ferrihydrite Fe2O3·xH2O Paramagnetic ∼40 125 - 500Goethite α-FeO·OH Antiferromagnetic 12.5 - 126 200 - 400Hematite α-Fe2O3 Antiferromagnetic 27 - 63 1360 - 1440Lepidocrocite γ- FeO·OH Paramagnetic 50 - 75 275 - 410Maghemite γ-Fe2O3 Ferrimagnetic 26000 - 44000 510 - 570Magnetite Fe3O4 Ferrimagnetic 39000 - 100000 275 - 371Quartz SiO2 Diamagnetic −0.58 573Water H2O Diamagnetic −0.90 100χ = mass-specific susceptibility.

41

Chapter 3

0

2000

4000

6000

8000

0 4 8 12

Sub 105 C weight loss [%]

Fe

[PP

M]

whiteorangered

o

Figure 3.6: A cross plot of the iron content from ICP-AES measurements and the free-water relatedweight loss during drying below 105 ◦C from the thermogravimetric analysis. The iron content is aproxy for the amount of iron oxides in the samples (see text). Although all samples were collected undersimilar conditions from nearby positions, small differences in grain size or time of exposure to air mayhave changed the amount of free water in the samples. This might explain the relatively large variationin weight loss %. The samples OS11-4 and OS11-5 were omitted from the diagram because their watercontent is strongly influenced by the proximal water table. The diagram shows a positive correlation,related to the amount of iron oxides in the sediment.

values show that the water occupies a maximum of around one third of the pore volume (Ta-ble 3.3), predominantly stored as capillary water in small pore bodies and pore throats. Theamount of water that can be retained as capillary or free water depends on the capillary pressureand pore shape (De Marsily, 1986). The amount of water that can be retained as adhesive water,which has a relative permittivity lower than that of free water (Roth et al., 1990), depends onthe specific surface of the sediment, but is small compared to the maximum amount of capillarywater.

It is shown in various studies using scanning electron microscopy (SEM) that iron-oxideminerals have a relatively rough surface texture relative to quartz grains (Frank, 1981; Smartand Tovey, 1981; Welton, 1984). Iron-oxide minerals like goethite appear to have a larger spe-cific surface than quartz, which allows for more adhesive water to be present per unit volume ofsediment. At a larger scale, the growth of iron-oxide minerals will create niches that can retaincapillary water (e.g., Berner, 1980). These features explain the correlation among iron-oxidecontent, free-water content and electromagnetic properties, as observed in the Ossendrecht sam-ples.

3.6.3 GPR

It is now obvious that, as long as the sediment contains water, variation in the abundance ofiron oxides in the sediment can have a significant influence on εr and, thus, on the electromag-netic signal. To understand the influence on the GPR signal, we constructed synthetic 2-D radarsections, for which we used εr values obtained during the laboratory TDR measurements (Ta-ble 3.5) and a µr of 1 (Table 3.2). Theoretically, a higher θ will result in more attenuation (α),but to keep the model elementary, we set α to 0.1 dB m−1. The velocities were calculated using

42


v = 0.133 m ns-1

v = 0.120 m ns-1

v = 0.109m ns-1

v =0.133 m ns-1

v = 0.109 m ns-1

Iron-oxideconfiguration

Sedimentarybedding

position [m]0 0.3 0.6 0.9 1.2 1.5

twt [

ns]

8

16

0

(c)

v = 0.120 m ns-1

v = 0.133 m ns-1

Sedimentarybedding

Boundingsurface

position [m]0 0.3 0.6 0.9 1.2 1.5

twt [

ns]

8

16

0

(a)

synthetic GPR imagevelocity modelfield example

v=0.120 m ns-1

v=0.133 m ns-1

v=0.109 m ns-1

v=0.133 m ns-1

v=0.115 m ns-1v=

0.130 m ns-1v=0.095 m ns-1

v=0.109 m ns-1

sand

fine sand

sand

position [m]0 0.3 0.6 0.9 1.2 1.5

twt [

ns]

8

16

0

(b)

white sand

orange sand

red sand

key:

0.2m

v = 0.109 m ns-1

Figure 3.7: Selected field examples of iron-oxide precipitates, velocity models and two-dimensionalsynthetic ray-trace GPR images: (a) precipitates that follow original bedding; (b) precipitates that cross-cut original bedding; and (c) discontinuities. The examples have dimensions of 1.5 m wide and 1 m high.The GPR synthetic radar sections were constructed using a Ricker wavelet and a 900-MHz frequency.

Equation (3.2). The results in Table 3.5 show that the velocities range from 0.095 m ns−1 forred (iron-oxide rich) fine sand with iron oxides to 0.133 m ns−1 for white sand. As iron-oxideprecipitation is a diagenetic process, the flow paths of the pore water play an important role inthe formation and morphology of iron-oxide rich zones. These flow paths are influenced by anumber of factors, such as sedimentary structures, discontinuities, and regional flow patterns.Based on these factors, we selected three field examples (Figure 3.7) that we translated intovelocity models using the data given in Table 3.5. These models, in turn, are input for the 2-Dray-trace modeling software.

The simplest situation is where the precipitates follow original cross-bedding or boundingsurfaces, as in Bristow (1995). The example in Figure 3.7a (based on site 5a in Figure 3.1)shows precipitates that were formed on a bounding surface. It is important to note that the top

43

Chapter 3

reflection (from the transition of orange to red sand) would not occur under a situation withoutan iron-oxide enriched layer. The bottom reflection (red to white sand) has a higher amplitudethan can be expected from the grain size variation at the bounding surface alone.

Figure 3.7b shows a situation where an iron-oxide band (made up of an orange upper partand a red lower part) cross-cuts depositional bedding (field example given in Figure 2.5). Thedepositional bedding is represented by a fine-sand layer, 0.2-m thick, within coarser sand. Thefine sand, containing more (capillary) water than the coarse sand, forms a low-velocity zone. Animportant feature to note is the non-horizontal morphology of the top and bottom reflections ofthis layer in the resulting radar section. The top reflection of the fine-sand layer dips graduallyto the right, whereas the bottom reflection is concave up. This feature is due to the velocitypull down by the material rich in iron oxides. The bottom reflection of the fine-sand layer losesamplitude towards the right, due to the low velocity contrast between white fine and normalsand. A real GPR section will show more complexity, owing to diffraction hyperbolas that willform at intersections.

An example of iron oxide that precipitated irregularly around a root system is given inFigure 3.7c. Here, the sedimentary bedding is horizontal. The velocity model is a simplifiedrepresentation of this situation. The synthetic radar section shows the difficulty in reconstructingthe original reflector morphology from the reflection configuration.

3.7 Conclusions

This study began with the premise that iron oxides might influence ground-penetrating radarperformance. Field data showed that iron oxides lower the electromagnetic wave velocityin sediments. For a set of samples containing a varying amount of precipitated goethite, allparameters that influence electromagnetic wave behavior were carefully analyzed. Magneticmeasurements showed that the relative abundance of iron oxides influences relative magneticpermeability consistently, but the variation is too small to be significant for electromagneticwave behavior. TDR measurements on bulk samples under field (humid) and dry conditions

Table 3.5: Electromagnetic wave velocities for a range of observed relative permittivities in field and lab,as well as estimates for fine sand (that was used for the GPR modeling) and some reference materials.

εr v [m ns−1]

Minimum observed field 4.0 0.150Maximum observed field 9.0 0.100White sand laboratory 5.1 0.133Orange sand laboratory 6.3 0.120Red sand laboratory 7.6 0.109White fine sand estimate 5.3 0.130Orange fine sand estimate 6.8 0.115Red fine sand estimate 10.0 0.095Air 1.0 0.300Saturated sand 25.0 0.060Water 80.0 0.034εr = relative permittivity;v = electromagnetic wave velocity.

44


showed that the electrical conductivity did not vary systematically with iron-oxide content.Using the dielectric mixing models we demonstrated that the iron oxides do not alter the rel-ative permittivity of the solid phase in the sediment. Thus, none of the three components ofelectromagnetic waves (µ, σ, and ε) is directly influenced by the iron oxides. However, the sub105 ◦C weight loss measured during the thermogravimetric analysis showed that the amountof iron oxide in the material is correlated with the volumetric water content and, thus, with thedielectric properties of the samples. The correlation is caused by the larger specific surface andcapillary-retention capacity of iron oxides like goethite, as compared with quartz grains. Ironoxides thus can have a profound influence on the relative permittivity and, consequently, on theGPR signal. Depending on the morphology and precipitation pattern, the iron-oxide zones mayeither exaggerate individual reflections, cause an increased attenuation, or complicate reflectionconfigurations and interpretation of GPR images.

45

Chapter 3

46

4

Influence of organic matter in soils onradar-wave reflection

AbstractOwing to the ability of organic matter to hold water, soils are excellent reflec-tors of ground-penetrating radar (GPR) signals. In this study, GPR profiles ofan eolian sedimentary succession are combined with textural, dielectric, andwater-retention characteristics to illustrate the influence of soil moisture onradar-wave reflection. Organic matter in this succession varies strongly, from<0.15% for clean sand to 7% for the most prominent soil. Water-retentioncurves show a complex relationship between suction potential (pF) and vol-umetric water content (θ). As a result of their uniform pore-size distribution,clean sand and weakly developed soils experience a sudden loss of water be-tween pF 1.5 and pF 1.8, going from saturated to almost dry conditions. Incontrast, the most prominent soil shows a more gradual decrease in θ with in-creasing suction potential. It follows that the dielectric contrast between cleansand and this soil increases abruptly above pF 1.5, reaches a maximum valueat field-capacity conditions, and then decreases slowly. Synthetic GPR im-ages for different suction potentials show that field-capacity conditions, whenreflection coefficients are high, are favorable for tracing one single soil hori-zon. Dry sediments are preferable when imaging widely spaced soils, whereassaturated sediments are best when imaging closely spaced soils.

This chapter is based on Van Dam, R. L., Van Den Berg, E. H., Van Heteren, S., Kasse, C., Kenter, J. A. M.,and Groen, K. (accepted). Influence of organic matter in soils on radar-wave reflection: sedimentological implica-tions. Journal of Sedimentary Research, scheduled for publication in the May 2002 issue, 72(3). Reprinted withpermission.

Chapter 4

4.1 Introduction

Ground-penetrating radar (GPR) has become a well-established technique to study the shallowsubsurface. Contrasts in electromagnetic properties, mainly governed by variations in watercontent, cause portions of transmitted electromagnetic energy to reflect to the surface. As or-ganic matter is capable of holding large quantities of water, soils act as excellent GPR reflectors(Van Overmeeren, 1994; Clemmensen et al., 1996; Barnhardt et al., 1998; Freeland et al., 1998;Van Heteren et al., 1998; Vandenberghe and Van Overmeeren, 1999). Although organic matteris common in sedimentary environments, most research on electromagnetic properties of clas-tic sediments has focused on properties such as grain size, density and water content; organicmatter is generally considered to be of secondary importance (Jackson, 1987; Wensink, 1993).A few studies of peat have used detailed analyses of organic-matter content and electromag-netic properties to improve the interpretation of GPR images (Worsfold et al., 1986; Warneret al., 1990; Theimer et al., 1994). Van Dam and Schlager (2000) quantified physical sedimentproperties to understand GPR reflection of mineral soils.

Doolittle and Collins (1995) show that variation in water content significantly affects GPRperformance in soils, but they do not discuss explanations for their observations. The purpose ofthe present research is to better understand the role of organic matter in radar-wave propagationand reflection under variable moisture conditions. To achieve this goal, a multidisciplinary fieldstudy of eolian dune deposits, located near the Dutch coast (Figure 4.1), was conducted withinthe HYDROSED (hydraulic characterization of sedimentary deposits) project. The sedimen-tary succession under study consists of beds containing clean windblown sands (organic-mattercontent <0.15%) and mineral soils with up to 7% organic matter. GPR data were collectedover an 11 × 16-m grid. Sediment characteristics were quantified along vertical profiles fromtwo trenches that were dug at the edge of the grid, following the GPR survey. Time-domainreflectometry (TDR) was used to measure small-scale variations in electromagnetic properties.Samples were collected to measure grain-size distributions and organic-matter content. In ad-dition, laboratory measurements of sediment water retention (soils and clean sands) were usedto calculate the dielectric properties for different water contents. Together, these data providedinput for the construction of synthetic radar traces, that in turn improve our understanding of theinfluence of organic matter and the inherent role of moisture on GPR performance. Improvedunderstanding on radar-wave reflection by organic matter is needed because GPR is used totrace positions and characteristics of soils in the subsurface for a number of reasons. In studiesof saturated fluid flow, soils are important low-permeable layers. Also, soils are used for datingsedimentary units and for stratigraphic and archeological studies.

4.2 Methods


The GPR technique is based on propagation and reflection of electromagnetic waves in the sub-surface. Fundamental properties that control the behavior of electromagnetic waves are dielec-tric permittivity (ε), electrical conductivity (σ), and magnetic permeability (µ), which togetherdefine the electromagnetic wave impedance (Z) (Von Hippel, 1954). A range of electromag-netic properties for common geologic materials can be found in Davis and Annan (1989) andin Van Heteren et al. (1998). For most natural sediments, variations in µ are insignificant (e.g.,

48

Influence of organic matter in soils on radar-wave reflection

4

8

12

X [m

]

0

16

Y [m]

Trench A

N

CMP 2

CMP 3 CMP 4

CMP 1

Tren

ch I

A

B

C

D

E

FG

H

1

2

3

4

5

6

7

11

AA

I

0

3

7

11

Lacquer peelTDR section

CMP measurementH GPR survey line

Trench A Trench I

0

16m11m

WSW

0 50km

N

Utrecht

Rhine

Study area

Line 1

Line A

(b)

(a)

Figure 4.1: (a) Study site with trenches. View is towards the westsouthwest. The inset shows the locationof the study area in The Netherlands. (b) Measurement grid.

49

Chapter 4

Daniels et al., 1988; Chapter 3), and the influence of σ on the electromagnetic signal velocity isnegligible over the entire GPR frequency range (Davis and Annan, 1989). In contrast, the rela-tive permittivity (εr = ε/ε0, where ε0 is the permittivity of free space) plays an important rolein both propagation and reflection of electromagnetic waves. Water has a relative permittivityof around 80, compared with air and quartz that have values of 1 and around 4.3, respectively.Therefore, the relative permittivity of bulk sediment is governed by its water content.

The propagation velocity of electromagnetic waves is essentially frequency independentbetween 100 and 1000 MHz, assuming σ <0.1 S m−1 (Davis and Annan, 1989; Powers, 1997).The propagation velocity (v) is then found by v = c0/

√εr, where c0 is the electromagnetic wave

velocity in vacuum (3 108 m s−1). Below 100 MHz, frequency dependence leads to higher εr

values and lower velocities. In this range, frequency dependence increases with water content.Attenuation of the electromagnetic signal is governed by σ and εr. In low-loss materials, the

original pulse amplitude (A0) decreases exponentially with depth (z) according to A = A0 e−αz,where the attenuation constant α = 0.5σ

√µ/ε (Theimer et al., 1994). This relationship is valid

only for homogeneous sediments without impedance contrasts. When a propagating electro-magnetic wave encounters an impedance contrast, part of the energy, proportional to the mag-nitude of change, is reflected. The reflection coefficient is given by RC = (Z2 −Z1)/(Z2 +Z1),where Z1 and Z2 are the impedances of the layers above and below the contrast, respectively.

For the synthetic modeling of GPR waves the program GPRMODV2 (Powers and Olhoeft,1995) was used. This program enables one-dimensional modeling of radar signals, includ-ing possible frequency dependence of electromagnetic properties. The program accounts forvariations in velocity and attenuation, caused by changes in water content and εr, at differentfrequencies. The result of this frequency dependence is pulse dispersion for frequencies below100 MHz.


The TDR method that was developed to characterize the conductivity and water content of soilsis based on the propagation of an electromagnetic signal along a probe. In addition to measure-ments of temporal variability, one can use TDR to construct vertical profiles of electromagneticproperties (Topp and Davis, 1985; Van Dam and Schlager, 2000). The relative permittivity (εr)is calculated from the travel time of the TDR signal in the sediment and from the probe length.Normally, the main factor that controls εr is the volumetric water content (θ) of sediment. Formineral soils with organic-matter contents up to 10%, θ can be found by substitution of εr inthe empirical relationship (Topp et al., 1980; Roth et al., 1992):

θ = −5.310−2 +2.9210−2εr −5.510−4ε2r +4.310−6ε3

r . (4.1)

The frequency-independent electrical conductivity (σDC) of sediment is found by estimating theattenuation of the TDR pulse (Giese and Tieman, 1975; Weerts et al., 1999).

4.2.3 Water-retention characteristics

Porous media in natural unsaturated conditions still retain water, which indicates that forcesprevent part of the interstitial moisture from draining. These so-called matric forces can besubdivided into adsorption forces and capillary forces (De Marsily, 1986). Adsorption is thestrong attraction between water molecules and the solid phase, creating a thin water film around

50


0 0.2 0.4 0.60

4

8

Volumetric water content

Suc

tion

pote

ntia

l

claysilty clay loamsandy loamsand

Figure 4.2: Examples of water-retention curves for sediment with different textural characteristics (afterBouma, 1977). Both position and shape of the curves are related to the pore-size distribution of thesediment.

sediment particles. The amount of adsorbed water retained by the sediment is limited, butincreases with the surface area of the solid phase. Capillary forces are the result of the surfacetension of a fluid relative to atmospheric pressure. Equating the forces on both sides of theair-fluid boundary shows that the capillary rise is inversely proportional to the pore radius (r).Matric forces thus exert suction (negative pressure relative to atmospheric pressure) on the porewater. This suction is usually expressed as the pressure head (h) or the suction potential (pF),where pF = log |h|. The suction potential is defined as pF = log(−pw/ρwg) (De Marsily,1986), where pw is the water pressure [N m−2], ρw is the water density [kg m−3], and g isthe gravitational acceleration [m s−2]. A suction potential of 0 to 1 is associated with saturatedconditions. Two terms are commonly used to compare between particular pF values in differentsamples (Ward and Robinson, 1990). The term ’field capacity’ is defined as the amount of waterremaining in sediment when drainage under the influence of gravity has ceased. The suctionpotential associated with field capacity depends on sediment type but is commonly in the rangebetween pF 1.5 and 2.5. Field-capacity conditions were present when the present field studywas conducted. The so-called ’wilting point’ is the minimum water content at which plants canextract water from sediment. To reach a pF value of 4.2, commonly associated with the wiltingpoint, gravitational drainage has to be assisted by evaporation and evapotranspiration.

The relationship between soil-water content and suction potential is frequently expressedin the so-called ’water-retention curve’ or ’capillary-pressure curve’. The shape of this curveis related to the pore-size distribution of the sediment (Figure 4.2). In general, soils contain awide range of interconnected pores of varying shapes and sizes. Those with wide entry channels(pore throats) will drain at low suction, whereas those with narrow channels will drain at higher

51

Chapter 4

suction. In a medium with a uniform pore-size distribution, like well-sorted sand, most of thepores drain in a narrow range of suction potentials. Sediments containing organic matter have awider range of pore sizes. Therefore, the pores drain in a wider range of suction potentials andassociated water-retention curves will have gentler slopes.

Experimental procedure — To measure the θ-pF relationship, essentially undisturbed coreplugs were saturated in the laboratory and subsequently drained over a pF range of 0.4 to 2.0.After allowing the samples to equilibrate with a specific suction for a week, the samples wereweighed and volumetric water content was calculated by θ = θgρb, where θg is the gravimetricwater content and ρb is the dry bulk density of the sediment. To determine porosity, the sampleswere dried at 60 ◦C for 24 hours, corresponding to a pF between 5.5 and 6.5. To constructmodels of water retention for the different units, the measured θ-pF pairs were fitted to theequation (Van Genuchten, 1980):

θ = θr +θs −θr

(1+(αvg|h|)n)mvg, (4.2)

where θr and θs represent the residual and saturated soil-water contents, respectively, αvg

[cm−1] and n are parameters to fit the shape and position of the curves, and mvg = 1− n−1.The computer program PFEREP (Waterloo, 1994) was used to make least-square estimates ofparameters αvg, which defines the position of the inflection point, and n, which determines theamount of curvature in the θ-pF relationship.

4.3 Sedimentology and stratigraphy of the test site

The test area was surveyed with a grid of 2-D GPR lines, using Sensors&Software pulseEKKO100 and pulseEKKO 1000 GPR systems with 25-, 100-, 225-, 450- and 900-MHz antennas(Tables 2.1 and 4.1). Subsequent to the GPR measurements, 2 trenches, each 3-m deep, weredug for detailed study and sampling of the sediment (Figure 4.1b). The water table is at adepth of around 3.5 m. Along vertical sections in the trench walls, TDR was used to obtaindetailed information on electromagnetic properties, and samples were taken for analysis oftextural properties (Table 4.2). A total of 49 core plugs were collected from the trench wallsfor laboratory measurements of water-retention characteristics (Figure 4.3a,b). The core plugsthat were sampled by pushing a metal ring in the sediment have walls of minimum thickness,so as to reduce sediment disturbance. In addition, samples of organic material were collectedfor 14C-dating. Lacquer peels were made for macroscopic study of textural characteristics.

The sediment is subdivided into 7 units (Figure 4.3a,b). Units 1, 3, 5, and 7 consist of cleanwindblown sand, whereas units 2, 4, and 6 are soils that contain varying amounts of organicmatter. Unit 1 is characterized by low-angle cross-stratification and horizontal lamination. Theminimum thickness of the unit is 0.5 m. Unit 2 is subdivided into an in-situ soil at the base ofthe unit and a mix of windblown sand and detrital organic matter at the top. The unit has anundulating lower boundary and is truncated at the top; in the northeastern part of the test siteit has been eroded completely. Here, a diffuse boundary exists between units 1 and 3. Severalrelics of human habitation like charcoal, pottery, and bones and teeth of domestic animals werefound in this unit. Unit 2 has a maximum thickness of about 0.9 m. Unit 3 is characterized bylow-angle cross-stratified sands and has a thickness that varies between 0.2 and 0.7 m. Unit 4 isan approximately 0.05-m-thin Ah-horizon soil that is continuous throughout the field site. Thesediments in unit 5, which has a thickness of 0.45 to 0.75 m, are characterized by horizontal

52


lamination. Unit 6 is a distinct soil with a uniform thickness of around 0.2 m. The soil has anundulating topography, such that the elevation of the top varies by about 0.55 m. The lower-lying sections of the soil (e.g. trench I, 8 m) contain more organic matter than the sectionsat higher elevations. The soil has Ah and Bw horizons in trench A; in trench I the soil is anaccumulation of organic matter in an Ah horizon. Towards the northwest in trench I, a thinsandy layer subdivides the soil into two parts. Unit 7 is characterized by high-angle cross-stratification, dipping towards the southeast, and has a minimum thickness of 0.2 m.

Ages of sampled organic material show that the sediment was deposited between around500 and 1425 years AD (Table 4.3). During a period of shoreline progradation between around5000 and 3000 14C-years BP (∼3750 to 1250 years BC), relatively low so-called Older Dunesformed on top of coastal barriers. From the Roman period (∼0 AD) onwards sand supply almostceased, causing barrier and dune development to diminish. Units 1 to 5 were deposited duringthis period. After about 1200 AD, a period of coastal erosion marked the onset of the formationof the so-called Younger Dunes (Jelgersma et al., 1970). Unit 6 marks this transition. Increasedsediment supply led to the development of up to 35-m-high dunes. Unit 7 represents the baseof the Younger Dunes.

Table 4.1: Details of GPR sections and CMP measurements. Locations can be found in Figure 4.1b.

Name X0 , Y0 Xmax , Ymax 25 MHz 100 MHz 225 MHz 450 MHz 900 MHz[m] [m] step size step size step size step size step size

0.5 m 0.2 m 0.1 m 0.05 m 0.025 m

11 0.00 , −1.00 16.00 , −1.00 - - - × -1 0.00 , 0.00 16.00 , 0.00 × × × × ×2 0.00 , 3.00 16.00 , 3.00 - × × × ×3 0.00 , 7.00 16.00 , 7.00 - × × × -4 0.00 , 11.00 16.00 , 11.00 - - × × -5 6.00 , 0.75 10.00 , 0.75 - - - - ×6 0.00 , 1.50 16.00 , 1.50 - - - × ×7 6.00 , 2.25 10.00 , 2.25 - - - - ×AA −1.00 , 0.00 −1.00 , 11.00 - - - × -A 0.00 , 0.00 0.00 , 11.00 - × × × -B 4.00 , 0.00 4.00 , 11.00 × × × × -C 8.00 , 0.00 8.00 , 11.00 × × × × ×D 12.00 , 0.00 12.00 , 11.00 - × × × -E 16.00 , 0.00 16.00 , 11.00 - - × - -F 2.00 , 0.00 2.00 , 11.00 - - - × -G 6.00 , 0.00 6.00 , 3.00 - - - × ×H 10.00 , 0.00 10.00 , 3.00 - - - × ×I 14.00 , 0.00 14.00 , 3.00 - - - × -CMP 1 6.00 , 0.00 - - × × × -CMP 2 0.00 , 6.00 - - - × × -CMP 3 6.00 , 4.00 - × - - - -CMP 4 8.00 , 1.50 - - - - - ×1

1 stepsize 0.020 m.

53

Chapter 4

06.1510 Y-position [m]

0

15

30

unit 6(top)

unit 4

unit 2 (top)

5

Line AA - 450 MHz

twt [

ns]

(c)

A7

2 046810

Scale in metersSSW NNE

A4

A1A2 A3

A6

Unit 3 -

Unit 5 -

A8

A9A10

A12

0

1

2

3

Trench A

A13

Unit 7 -Unit 6 -soil

Unit 4 -

Unit 2 -

Unit 1 -

low-angle cross stratification

horizontal lamination

high-angle cross stratification

soil

soil

low-angle cross stratificationand horizontal lamination

3

C1

C4

(a)

Dep

th [m

]

7

A5

Figure 4.3: Trench wall panoramas, showing sedimentary units and locations of measurement and sam-pling points for (a) trench A, and (b) trench I. 450-MHz GPR sections from (c) line AA, and (d) line 11allow correlation with the pictures. The GPR data were imaged with the GPRMODV2 program (Powersand Olhoeft 1995).

Table 4.2: Details on vertical sections in the trench walls that were used to obtain measurements ofelectromagnetic properties (TDR) and samples for grain-size and organic-matter analyses. Locations areindicated in Figures 4.1b and 4.3a,b.

Trench Section X , Y Length ΔY[m] [m] [m]

A TDR 3 0.00 , 6.15 2.875 0.025A TDR 7 0.00 , 6.60 2.825 0.025A Sampling 0.00 , 6.40 2.350 0.025I TDR 1 8.35 , 0.00 2.750 0.025I TDR 8 8.80 , 0.00 2.750 0.025I Sampling 8.60 , 0.00 2.650 0.050

54


0

1

2

3

0 4 6 12 142 10

Scale in metersESE WNW

I2

I3

I5

I7

I19

I16I17

I18

I22

truncationlevel

I23 I21

I21

I26I25

I27 I28

I20

Trench I

Unit 7

Unit 6

Unit 4Unit 5

Unit 3Unit 2

Unit 1

I22

I6b

C2

C10

C11

I1

(b)

1 8

0 8.35 15X-position [m]

0

15

30

unit 2/4

5 10

unit 6 (top)

unit 6 (bottom)

unit 4

Line 11 - 450 MHz

twt [

ns]

(d)

undisturbed core plug for mea-surement of water retentionA9 samples for textural

characteristicssample for14C dating

C13 TDRmeasurements

outline oflacquer peels

Legend

Table 4.3: Carbon-14 and calibrated ages of organic matter sampled from the trench walls (see Figure 4.3for locations). The cow tooth representing the top of unit 2 provides a maximum age. Originally, it mayhave been deposited at a nearby location. Ages were calibrated using Stuiver et al. (1998).

Code Trench Unit Sample Description 14C age Calibrated age[years BP] [years AD]

GrN-24489 A 2 C1 Charcoal 1560 ±80 475 - 530GrA-13881 A 2 top C4 Cow tooth 1360 ±40 660GrN-24490 I 4 C2 Wood 1090 ±40 975GrN-24746 I 6 bottom C10 Bulk organic matter 790 ±80 1260GrN-24492 I 6 bottom C10 Bulk organic matter 760 ±20 1270GrN-24493 I 6 top C11 Wood 500 ±40 1425

55

Chapter 4

4.4 Results


The 450-MHz GPR lines in Figures 4.3c and 4.3d correspond to the trench walls. On GPRline 11, the top of unit 6 is marked by a sharp, high-amplitude reflection (Figure 4.3d). Thebottom reflection of unit 6 is less distinct, possibly because of multiple reflections within theunit and of interference with diffraction hyperbolas originating in the top of the unit. Unit 6has a larger ’time-thickness’ right of meter 5 where higher organic-matter and water contentsinduce a lower velocity. The result is velocity pull down for the bottom reflection of unit 6 andfor all lower reflections. Unit 4 gives a good reflection in the left marginal area of line 11. In theright marginal area, units 4 and 2 produce one combined reflection. In the central part of thisGPR line, the reflection intensity below unit 6 is lower than elsewhere; the clean sand seemsalmost transparent to GPR signals. This fact may indicate a high scattering loss from the uneventop of unit 6 and from non-horizontal incidence to the high-angle cross-bedding in unit 7. InGPR line AA (Figure 4.3c), both the noise level and the signal level are lower than those inline 11, for unknown reasons. No clear bottom reflection for unit 6 is present. Nevertheless, thelower-lying units 2 and 4 give low-amplitude but good reflections.

4.4.2 Sediment description

The average grain size of the studied sediment is around 260 µm (medium sand) for both verticalsections (Table 4.4). The soils have a slightly finer grain size than the clean sand (Figure 4.4),which is mainly expressed by an increase in silt content (Table 4.4). Clay content is below 1%for both clean sand and soils. Within the observed textural range the grain size does not affectthe TDR response much; for the clean sands εr remains at a baseline value of around 4 in allsections (Figure 4.4). The sharp increase near the base of the trench is due to the proximity ofthe water table. The average percentage of organic matter is well below 1. In units 2, 4, and 6,this percentage is higher (Figure 4.4, Table 4.4). Unit 6 peaks at 2.5% organic matter in trench Abut reaches 5% in trench I, showing εr values of up to 26. Within unit 6, organic-matter contentvaries not only laterally but also vertically (Figure 4.4). Units 2 and 4 contain less organicmatter than unit 6. Unit 4 lacks a response in the TDR measurements in trench A, whereas intrench I a minor excursion from the baseline value is visible at 1.5-m depth (Figure 4.4). The

Table 4.4: Textural characteristics of the samples from the vertical sections in trenches I and A. Clayand silt are defined as the weight fractions smaller than 2 µm and between 2 and 62 µm, respectively.

Trench N Grain size [µm] Clay content Silt content Organic matterMean Sd % Sd % Sd % Sd

I - clean sand 40 264.96 21.91 0.48 0.06 0.73 0.22 0.46 0.21I - organic units 13 238.74 19.48 0.64 0.10 1.74 0.70 2.21 1.64I - total 53 258.53 24.03 0.52 0.10 0.98 0.58 0.89 1.11A - clean sand 78 264.43 18.43 0.61 0.08 0.80 0.17 0.43 0.18A - organic units 14 243.46 19.02 0.81 0.14 1.97 0.83 1.27 0.64A - total 92 261.34 19.87 0.64 0.11 0.97 0.54 0.56 0.42N = number of samples; Sd = standard deviation.

56


0

0.5 1

1.5 2

2.5 3

0

0.5 1

1.5 2

2.5 3

0

0.5 1

1.5 2

2.5 3

0

0.5 1

1.5 2

2.5 3

0

0.5 1

1.5 2

2.5 3

0

0.5 1

1.5 2

2.5 3

010

2030

010

2030

02

46

200

250

300

350

00.

20.

40

0.2

0.4

Rel

ativ

ep

erm

itti

vity

1R

elat

ive

per

mit

tivi

ty 8

Org

anic

mat

ter

[%]

Gra

in s

ize

[μm

]V

olu

met

ric

wat

erco

nte

nt

1V

olu

met

ric

wat

erco

nte

nt

8

Depth [m]

TD

R3

TD

R7

Uni

t 7

Uni

t 4

Uni

t 5

Uni

t 3

Uni

t 2

Uni

t 1

Uni

t 6

0

0.5 1

1.5 2

2.5 3

0

0.5 1

1.5 2

2.5 3

0

0.5 1

1.5 2

2.5 3

0

0.5 1

1.5 2

2.5 3

0

0.5 1

1.5 2

2.5 3

0

0.5 1

1.5 2

2.5 3

02

46

010

2030

010

2030

200

250

300

350

00.

20.

40

0.2

0.4

Org

anic

mat

ter

[%]

Rel

ativ

ep

erm

itti

vity

3R

elat

ive

per

mit

tivi

ty 7

Gra

in s

ize

[μm

]V

olu

met

ric

wat

erco

nte

nt

3V

olu

met

ric

wat

erco

nte

nt

7

Depth [m]

Tre

nch

ITD

R8

TD

R1

Uni

t 7

Uni

t 4

Uni

t 5

Uni

t 3

Uni

t 2

Uni

t 1

Uni

t 6

Tre

nch

A

Figure 4.4: Variability of textural and electromagnetic properties with depth in trenches A and I. Picturesof the trench walls that show the locations of TDR sections, correlate the sediment profile with themeasurements. The textural characteristics (grain size and organic-matter content) correlate well withthe variation in relative permittivity and volumetric water content.

57

Chapter 4

0

5

10

15

20

25

0 1 2 3 4 5 6

N = 274

Rel

ativ

e pe

rmitt

ivity

Organic-matter content [%]

Figure 4.5: Variation of relative permittivity versus organic-matter content. The quality of the correlationis obscured by various factors. Water that stagnates atop and hangs under layers of low permeabilitycauses εr to increase in the sediment above and below soils (Figure 4.4). Undulating soil horizons, suchas unit 4 at a depth of about 1.5 m in trench I (Figure 4.4b), cause occasional mismatches between peaksin organic content and εr.

TDR response of unit 2 mimics the organic content well, leading to maximum εr values of 6 and8 in trenches A and I, respectively. The electrical conductivity (σDC) ranges from 0.0045 S m−1

for the clean sand to 0.0170 S m−1 for the soils. Substitution of the measured extremes forσDC and εr in equations for electromagnetic wave impedance and attenuation shows that thevariation in σDC is of minor importance.

The relative permittivity thus shows a pattern of variation correlated predominantly to orga-nic-matter content (rather than to grain size), as is shown by the relationship in Figure 4.5(R2 = 0.71). The quality of the correlation is somewhat obscured by water that stagnates atoplow-permeable soils. As a result, εr increases above soils, as is the case for unit 6 in bothtrenches (Figure 4.4). Also, below these soils εr tends to decrease more gradually than theorganic-matter percentage.

4.4.3 Water-retention curves

Of the 49 undisturbed samples that were used to construct water-retention curves, 42 werecollected from the clean-sand units, 5 from unit 2 and 2 from unit 6. The clean-sand units1, 3, 5 and 7 were merged into one group, as their characteristics appeared to be very similar(Table 4.5). Measured and modeled soil-water contents, presented in Table 4.6 along withthe optimized model parameters (Van Genuchten, 1980), allow construction of water-retentioncurves for the clean-sand group and the 2 soils (Figure 4.6a). The clean sand has a volumetricwater content of 0.43 at complete saturation. As suction potentials are increased from 0 to 1.5,

58


0

1

2

3

4

5

0 0.1 0.2 0.3 0.4 0.5 0.6


Suc

tion

pote

ntia

lunit 1,3,5,7

unit 2

unit 6

unit 1,3,5,7

unit 2

unit 6

0 10 20 30 40 500

1

2

3

4

5

Relative permittivity

(b)

Suc

tion

pote

ntia

l

(a)

Figure 4.6: (a) Water-retention characteristics for 3 groups based on 49 samples from 6 units. Averagevalues from the laboratory measurements are shown as data points. Best-fit water-retention curves areshown as solid lines. The decrease in water content going from full saturation at pF 0 to pF 0.40 is mostlikely due to a systematic procedural error during the experiment. The mismatch between the measuredvalues and the modeled curves for clean sand and unit 2 over the pF range from 0 to 1.5 is also found inother experimental studies of water retention in sand and silt (Fredlund and Xing 1994). (b) Curves basedon Figure 4.6a and Equation (4.1), showing the relation of suction potential versus relative permittivityfor the 3 groups. Depending on the amount of organic matter in the sediment the water content anddielectric properties react in different ways to changing suction potentials, as can be seen from the curveshapes and positions. Relative permittivity values, shown as data points, at pF 1, 1.866 and 4 were usedto construct synthetic radar images.

59

Chapter 4

water content decreases only slightly. Above pF 1.5, water content drops rapidly to 0.066 andthen stabilizes again at suctions higher than 1.8. The sharp drop in water content is characteristicof sediment with a uniform pore-size distribution, and is frequently found in well-sorted sandand sandstone (Van Genuchten, 1980; Fredlund and Xing, 1994). The shape of the curve forunit 2 is comparable to that of the one for clean sand. However, water contents are significantlyhigher, which reflects the presence of organic matter in unit 2. The two samples from unit 6have considerably higher volumetric water contents throughout the entire pF range. Their watercontent at saturation is 0.59. Between suction potentials of 0 and 1.5, the shape of the water-retention curve is not different from that of the other 2 curves, but at suctions higher than 1.5there is a gradual decrease in water content unlike the drop-off found for unit 2 and the clean-sand group.

The modeled θ-pF relationship was used as input in Equation (4.1) in order to show therelationship between suction potential and relative permittivity (Figure 4.6b). The curves haveabout the same shape as those in Figure 4.6a, but at suctions below 1.5 the curves of the threegroups are widely spaced, whereas at suctions above 3 the curves are within a small range ofεr values. At pF 1.0 the difference in εr between clean sand and unit 2 is about 7, whereas thedifference between unit 2 and unit 6 is around 13. At a suction potential of 1.8, the differencein εr between clean sand and unit 2 is minimal, whereas unit 6 is highly distinct from the otherunits. At pF 4.0, the differences in θ and εr are small among all three units.

Implications — The differences in curve shape, maximum water content, and relative per-mittivity among the different units have several important implications. One effect is that con-trasts in dielectric properties between units will differ significantly according to the moistureconditions (Figure 4.6b). At suctions above pF 3, average velocity will be high and attenua-tion low but the impedance contrasts will be small. At suctions below pF 1.5, the impedancecontrasts will be moderate but now with low average velocity and a higher attenuation. AroundpF 1.8, unit 6 exhibits large impedance, attenuation and velocity contrasts with the other units.For these three pF situations, large differences in response of the units to GPR signals can beexpected. Another effect will be that for clean sand and unit 2, a change in suction potentialbetween 1.5 and 1.8 results in a relatively large change in volumetric water content. Smallvariations in suction over this range might affect the GPR signal significantly.

The correlation between organic content and dielectric properties of a sample allows con-struction of vertical models of the contrast parameter εr. The validity of the dielectric valuesused for these models can be verified by calculating the volumetric water contents for differentunits using Equation (4.1) and then substituting these values in Equation (4.2) with the appro-priate parameters. The combination of θ values (measured with TDR) and the water-retention

Table 4.5: Average textural characteristics of the 49 samples used for the water-retention measurements.Clay and silt are defined as the weight fractions smaller than 2 µm and between 2 and 62 µm, respectively.

Trench N Grain size [µm] Clay content Silt content Organic matterMean Sd % Sd % Sd % Sd

Units 1,3,5,7 42 262.02 14.91 0.60 0.08 1.07 0.38 0.24 0.30Unit 2 5 246.57 2.70 0.80 0.10 1.90 0.68 0.92 0.17Unit 6 2 212.78 3.61 0.86 0.12 3.04 0.55 6.30 0.99N = number of samples; Sd = standard deviation.

60


curves gives a narrow range of suction potentials and shows that the field capacity at the testsite was about pF 1.866 (Table 4.7). Conversely, the water-retention curves can be used to con-struct contrast-parameter models for different suction potentials. These models in turn allowconstruction of synthetic radar measurements and comparison with actual GPR traces from thetest site.

4.5 Synthetic GPR modeling

Using the data from Figure 4.6b, vertical models of relative permittivity were constructed forthree different suction potentials (Table 4.8, Figure 4.7). Unit 1 is assumed to be of infinitedepth. Unit 4, that lacks any measurements of water-retention characteristics, is assigned thesame properties as unit 2. Simplifying the gradual changes in εr around the soils (Figure 4.7b)by blocks with constant values and sharp transitions makes interpretation of synthetic imagesmore straightforward, but results in a loss of information. The amount of information missed isgoverned by the ratio between the transition-zone width (W ) and the wave length (λ) of the radarsignal. Transitions are best imaged when W/λ is below 0.3 (Annan et al., 1991). Higher ratiosinduce signal dispersion and a decrease in reflection amplitude. For a velocity of 0.12 m ns−1,a transition zone of 0.1 m, as observed for unit 6, can be imaged with all antennas except the900-MHz one. This allows for the use of blocks with sharp εr transitions in the models.

Ricker transmission wavelets with center frequencies of approximately 100 and 450 MHzwere used for the forward modeling (Figure 4.8). GPR synthetics at field-capacity conditionsshow that the 20-ns-wide 100-MHz pulse is too coarse in resolution to image the different soils(Figure 4.9). In field measurements, direct air and ground waves and interference from mul-tiples would obscure the trace even more. The 450-MHz synthetic clearly images the polarityreversal for the top and bottom reflections of unit 2 and 6, whereas for unit 4 the top and bottomreflections have merged. The reflection amplitude for units 2 and 4 is small.

Table 4.6: Mean volumetric water content measured for different suction potentials in 49 undisturbedsamples. Modeled values (Van Genuchten, 1980) and optimized model parameters are given in italics.

Unit θs Suction potential (pF) θr αvg n r[m3m−3] 1 1.5 1.8 2.0 [m3m−3] [cm−1]

Unit 1,3,5,7 0.421 0.387 0.366 0.066 0.053 - - - -0.400 0.400 0.371 0.061 0.047 0.045 0.024 6.920 0.99

Unit 2 0.495 0.441 0.409 0.116 0.096 - - - -0.460 0.460 0.420 0.117 0.095 0.095 0.025 6.840 0.98

Unit 6 0.592 0.535 0.484 0.408 0.351 - - - -0.550 0.537 0.481 0.406 0.353 0.150 0.023 1.707 0.97

θs = saturated soil-water content; θr = residual soil-water content; r = correlation coefficient.

61

Chapter 4

(a)

(b)

(c)

4.2

4.2

4.2

4.2

6.3

6.3

23.9

0

1

2

3

pF = 1.866

pF = 1

25

25

25

25

32.2

32.2

44.5

0 10 200

1

2

3

30 40

1

2

3

4

5

6

7

pF = 4

3.6

3.6

3.6

3.6

5.9

5.9

8.7

0

1

2

3

1

2

3

4

5

6

7

1

2

3

4

5

6

7

Dep

th [m

]D

epth

[m]

Dep

th [m

]

Uni

tU

nit

Uni

t

0 10 20 30 40

0 10 20 30 40


Figure 4.7: Simplified vertical models of relativepermittivity for TDR section 1 (Figure 4.4) at (a)saturated conditions with pF 1, (b) field-capacityconditions with pF 1.866 (the TDR field data areshown for reference), and (c) low-moisture condi-tions with pF 4. Unit 4, having comparable textu-ral properties as unit 2, is given the same dielectricproperties as unit 2. GPRMODV2 uses input forthe high- and low-frequency limits of relative per-mittivity; the εr values shown are those at the centerfrequency.

62


(a)

(b)0 40 0 400200

0 8 0 1200800

Frequency [MHz]Time [ns]

20

4

103

Power spectrumTransmitted pulse

464

Figure 4.8: Pulse shape and frequency distribution of Ricker transmission wavelets used in GPR mod-eling; (a) 100 MHz and (b) 450 MHz. With increasing antenna frequency, a smaller part of the powerspectrum is in the zone below 100 MHz were dielectric properties behave frequency dependent.

6T

6B

4

2T

2B

0

20

40

60

twt [

ns]

450 MHz

100 MHz

Unit 7

Unit 6

Unit 5

Unit 4

Unit 3

Unit 2

Unit 1

Figure 4.9: GPR synthetic images for 100 and 450 MHz at field-capacity conditions (pF 1.866) andpicture of trench I between TDR sections 1 and 8. The reflection events in the 450-MHz trace areindicated with their unit number and a top or bottom indication. The data are plotted with a constant gainof 68 decibel.

63

Chapter 4

Figure 4.10 compares 450-MHz synthetic images for the three different models as shown inFigure 4.7. Important observations can be made with respect to reflector spacing and reflectionamplitude. Because of higher electromagnetic wave velocities for dry sediment the reflectionsare squeezed together at increasing suction potential (Figure 4.10). The transition from pF 1to field-capacity conditions is the most distinct in this respect, as a result of the draining of theclean sand and units 2 and 4 between pF 1.5 and 1.8 (Figure 4.6). The transition from pF 1.866to pF 4 causes the top and bottom reflections from unit 6 to squeeze. Multiples, present in fieldmeasurements, will problematize the interpretation of this trace at a pF of 1.866 and 4. Theamplitude of reflections gives information about signal attenuation and reflectivity of the layers.At field-capacity conditions (pF 1.866), the amplitudes of the unit-2 and unit-4 events are verylow. This is caused by the high permittivity contrast between unit 6 that still holds a lot ofwater and clean sand that is almost dry at this suction potential (Figure 4.6). As a result of thehigh reflection coefficient a relatively small amount of energy will pass unit 6 (Table 4.8). Forboth saturated (pF 1) and dry conditions (pF 4) the unit-6 reflections have significantly loweramplitudes, whereas the reflection amplitudes for units 2 and 4 are higher (Figure 4.10). Thelower reflectivity of unit 6 holds a larger part of the transmitted energy available for imagingunits 2 and 4. Reflection amplitudes can not directly be correlated with reflection coefficients(RC). For the unit-4 transition RC differs by a factor 2 between pF 1 and pF 4 (Table 4.8).However, at pF 4 the reflection amplitude for unit 4 is more than doubled compared with thepF 1 amplitude (Figure 4.10). This effect is caused by a significantly higher attenuation insaturated sediment.


Figure 4.11 shows the variation in reflection coefficient (RC) with suction potential for a vague(unit 2) and a prominent soil (unit 6). The contrast between clean sand and unit 6 (soil with 6to 7% organic matter) is highly variable. In saturated conditions RC is more or less constant.The dip to lower RC values at a suction of around pF 1.4 is somewhat overestimated, causedby the mismatch between measured and modeled water-retention characteristics for clean sand(Figure 4.6). Going from pF 1.5 to 1.9 the contrast nearly triples, followed by a gradual halv-ing towards pF 4 (Figure 4.11). For the unit-2 (and 4) transition the RC variation is smaller(Figure 4.11). The sharp drop in water content going from pF 1.5 to 1.8 (Figure 4.6) causes asmall increase in reflection coefficient.

Table 4.7: Volumetric water content values from the TDR measurements (Figure 4.4) were substitutedin Equation (4.2) to find the average suction potential (pF) at field conditions.

Unit εr θ [m3m−3] pF

1,3,5,7 4.21 0.06 1.882 7.19 0.13 1.826 22.63 0.38 1.90Average 1.866εr = relative permittivity;θ = volumetric water content.

64


2T

2B

6T

6B

4

pF 1

.866

2T

2B

6T

6B

4

pF 4

2T

2B

6T

6B

4

0 20

40

60

80

twt [ns]

Unit

7

Unit

4

Unit

5

Unit

3

Unit

2

Unit

1

Unit

6

pF 1

Figure 4.10: GPR synthetic images at 450 MHz for different suction potentials as shown in Figure 4.7.Picture of trench I between TDR sections 1 and 8 is shown for reference. The reflection events in thetrace at pF 1 are indicated as in Figure 4.9. The data are plotted with a constant gain of 68 decibel.

65

Chapter 4

clean sand - unit 2

clean sand - unit 6

0 0.1 0.2 0.3 0.40

1

2

3

4

Reflection coefficient

Suc

tion

pote

ntia

l

Figure 4.11: Relationships between reflection coefficient (RC) and suction potential (pF) for the transi-tion from clean sand to unit 2 and for the transition from clean sand to unit 6. Symbols mark the suctionpotentials used for the GPR modeling (Figures 4.7 and 4.10). The dip to lower RC values at pF 1.4 forthe unit-6 contrast is exaggerated due to erroneous modeling of clean-sand water-retention characteris-tics (Figure 4.6). The dip to lower RC values at suctions around 1.6 for the unit-2 contrast is probably amodeling artefact (Figure 4.6).

Table 4.8: Water contents, relative permittivity, and reflection coefficients for the different units at satu-rated conditions (pF 1), field-capacity conditions (pF 1.866), and low-moisture conditions (pF 4). Bothθ and εr decrease with decreasing organic-matter percentage and with increasing pF . For the transitionfrom clean sand to unit 6, RC is largest at field conditions. For the clean sand to unit-2 transition RCincreases with suction potential.

Suction Volumetric water content (θ) Relative permittivity (εr) Reflection coefficient (RC)Potential Unit Unit Unit Unit Unit Unit Clean sand Clean sand

1,3,5,7 2 6 1,3,5,7 2 6 to unit 6 to unit 2&4

1 0.400 0.460 0.537 25.0 32.2 44.5 0.14 0.061.866 0.057 0.105 0.389 4.2 6.3 23.9 0.41 0.104 0.045 0.095 0.159 3.6 5.9 8.7 0.22 0.12

66


A single GPR reflection will never be unique; it is thus not possible to directly determinethe type of sediment from the reflection characteristics in one trace. However, the comparisonof measurements from varying moisture conditions will outline soil reflectors and make labo-rious and costly drillings unnecessary. The reflection amplitude for soils will show a largertemporal variation than the reflection amplitude of sedimentary structures. In temperate regionsGPR measurements during wet conditions (directly after a period of rain) should be comparedwith measurements under field-capacity conditions (pF 1.866). For soils with small amounts oforganic matter the behavior is not markedly different from clean sand. In arid regions were suc-tion potentials above 2.5 or 3 are more common than saturated conditions, temporal amplitudedifferences between moist (pF 1.5 to 2.5) and dry (pF > 3) conditions should be compared tooutline prominent soils. However, in arid climates soils usually contain only small amounts oforganic matter. As the shape of the water-retention curve for soils with few organic matter lookslike the clean sand one, it is difficult to outline these soils without bore hole control.

Depending on the goal when studying soils, measurements under different moisture condi-tions are preferable. When one soil is present in the sediment and the aim is to track the extentof this horizon it is best to measure under field-capacity conditions when the dielectric contrastis large. The result is a reflection of high amplitude that can be traced easily. When focus is ona series of stacked soils, field-capacity conditions are less preferable for GPR research as thelarge dielectric contrasts of soils cause too much electromagnetic energy to reflect. Instead, itis better to measure when reflection coefficients are small. When the various soils are closelyspaced it is best to measure in saturated sediment such that reflected pulses do not overlap.When the stacked soil layers have a wider vertical separation it is best to measure under dry,low-attenuative conditions.

GPR tracing of the position and characteristics of soils in the subsurface is important for anumber of reasons. Soils are used for dating sedimentary units, for making (chrono)stratigraphiccorrelations, and for the reconstruction of sedimentary systems. In studies of saturated fluidflow, soils are important low permeable layers (aquitards). This study improves the interpreta-tion of GPR images of soils and allows for a more specific use of the technique. In addition, itprovides an explanation of the complex relation between organic-matter content and variationsin water content and their control on GPR signals.

67

Chapter 4

68

5

Radar reflections from sedimentarystructures in the vadose zone

AbstractGround-penetrating radar (GPR) is considered a suitable technique to imagesedimentary structures in the vadose zone because small texture variationslead to changes in capillary pressure and water content that, in turn, causereflection of electromagnetic waves. To study exactly how GPR reflectionsare generated by sedimentary structures, GPR profiles of an eolian sedimen-tary succession are combined with measurements of textural, electromagnetic,and water-retention characteristics from a trench. Time-domain reflectometryindicates that small texture variations in the high-angle dune sediment are as-sociated with changes in water content. Synthetics show that these changescause clear GPR reflections. In an experimental approach to estimate theradar response of structures below the wave resolution, variations in grain-size distribution and porosity in a thin section were used to reconstruct water-retention curves and impedance models of the thinly layered sediment. Syn-thetic radar records calculated from the impedance models show that reflec-tions from the studied sub-centimeter-scale structures are composites of inter-fering signals. Although low-amplitude interfering signals will commonly beoverprinted by more prominent reflections, they may cause reflection patternsthat change with frequency and do not represent primary bedding.

This chapter is based on Van Dam, R. L., Van Den Berg, E. H., Schaap, M. G., Broekema, L. H., and Schlager,W. (submitted). Radar reflections from sedimentary structures in the vadose zone. In: Ground penetrating radarin sediments: applications and interpretation (Ed. by C. S. Bristow & H. M. Jol), Geological Society SpecialPublication.

Chapter 5

5.1 Introduction

Ground-penetrating radar (GPR) is frequently used to image sedimentary structures for recon-struction of depositional history or for reservoir characterization. It is generally understood thatchanges in texture cause variations of water content (e.g., Mishra et al., 1989; Hanninen, 1992).Because water has dielectric properties that are highly contrasting with those of air and quartz,the water content of sediment governs the behavior of radar waves. Thus, water-content varia-tions that are caused by texture variation in sedimentary structures lead to changes in dielectricproperties. As radar signals reflect from boundaries between layers with different electromag-netic properties, GPR can be used to image sedimentary structures (e.g., Huggenberger et al.,1994; Kowalsky et al., 2001).

Studies that try to quantify the relationship between textural properties, water content andelectromagnetic wave propagation usually consider a large range of grain and pore sizes (e.g.,Hubbard et al., 1997; Endres, 2000). Frequently used regression functions (Gupta and Larson,1979; Saxton et al., 1986) and the more sophisticated physico-empirical model of Arya andParis (1981) relate water content to bulk density and to percentages of sand, silt, clay andorganic matter. Knoll and Knight (1994) presented a dielectric mixing model that includesclay-volume fraction. Sutinen (1992) quantified relations between grain size, water content,and dielectric properties. He found that the percentage of clay and silt had a large effect on thedielectric properties but observed no correlation between dielectric properties and the grain sizeof sand and gravel.

Eolian sediment, discussed in the present study, commonly has a very narrow range of grainsizes and usually contains only grains in the sand fraction. Still, eolian deposits produce clearGPR reflections (e.g., Bristow et al., 2000a; Van Dam, accepted). For these rather uniform sed-imentary facies, important questions about reflection of radar signals remain unsolved. Withrespect to GPR, it is not known in great detail which textural characteristics (e.g., grain size,sorting, packing, grain shape, porosity) control the actual reflection. Also, the minimum re-quired variation in these textural characteristics to generate a reflection is unclear. Furthermore,little is known about the contribution of reflection and signal interference from layering smallerthan the vertical resolution limit, commonly assumed to be a quarter wavelength, λ/4 (e.g.,Huggenberger, 1993). Typical values for λ/4 in moist sand (v = 0.12 m ns−1) range from 0.3 to0.033 m for 100- to 900-MHz frequencies, respectively (Table 2.1).

Interference of radar reflections from small-scale electromagnetic-property variation in ge-ologic material and sediment is very common (e.g., Clement et al., 1997) but has never beenstudied in detail. Few modeling studies on this issue have been performed. In seismic reflectionthis problem has gained much more attention from both fields of interpretation (e.g., Roksandic,1995) and modeling (e.g., Kallweit and Wood, 1982). Gochioco (1992) shows that layers as thinas one-fortieth the width of the dominant wavelength are detected. However, thin sediment lay-ers are commonly not isolated and are often more narrowly spaced than vertical resolution,such that recorded reflections are composites of several interfering signals (e.g., Mayer, 1980;Knapp, 1990; Bracco Gartner, 2000). As a result, the seismic response from thinly layeredsediment may not represent distinct geologic horizons (Mayer, 1979).

To address the above questions, we selected a study area in eolian dune deposits near theDutch coast, south of Katwijk aan Zee (Figure 5.1). The deposits consist of units with cleanwindblown quartz sand, separated by soils (Chapter 4). The upper unit, characterized by high-angle cross-stratification, was used for this study. The site was surveyed using a grid of GPRlines. Next, trenches were dug that allowed us to study and sample the sediment and to measure

70

Radar reflections from sedimentary structures in the vadose zone

4

0

0

3

11

Y [m

]

7

CMP 1

2

6

1

5

7H CMP 4

C

G

Tre

nch

A

Lacquer peel

TDR section

CMP measurement

H GPR survey line

N

Trench I

X [m]

(a)

(b) 81216

0

16m

11m

WSW

Outline of GPR grid Detailed study area

Trench ITrench A

0 50km

N

Utrecht

Rhine

AmsterdamKatwijkaan Zee

Figure 5.1: (a) Study site with trenches. The inset shows the location of the study area in The Nether-lands. The dashed lines mark the detailed survey area, imaged with 450- and 900-MHz antennas. (b)Measurement grid.

71

Chapter 5

electromagnetic properties. Lacquer peels were collected for macroscopic study of sedimentarystructures and samples for grain-size analyses were taken from a vertical transect. Also, coreplugs were collected to determine water-retention characteristics of the sediment in the labo-ratory. Time-domain reflectometry (TDR) was used to obtain direct and detailed informationon the electromagnetic properties of the sedimentary structures. Finally, box cores were col-lected for microscopic texture analysis in thin sections. The small-scale textural informationfrom the digitized thin sections was used to estimate water retention and dielectric propertiesperpendicular to the sedimentary bedding. The TDR measurements and thin sections were usedto construct models of electromagnetic wave impedance and to calculate synthetic radar traces.The TDR-based synthetic radar traces were used to study signal interference. The thin-sectionbased model serves as a more theoretical approach to improve understanding of propagationand reflection of GPR waves within and across sedimentary structures.

5.2 Water retention

The primary reason that GPR can image sedimentary structures is that the water content of un-saturated sediments is a function of sediment properties that vary most prominently perpendic-ular to depositional layering. Porous media in natural unsaturated conditions retain some water,which indicates that forces prevent part of the interstitial moisture from draining. These so-called matric forces can be subdivided into adsorption and capillary forces (De Marsily, 1986).Adsorption is the strong molecular attraction between water and solids that creates a thin waterfilm around sediment particles. The amount of adsorbed water depends on the specific-surfacearea of the solid phase. Because adsorbed water has a low relative permittivity (Roth et al.,1990; Saarenketo, 1998) and because the specific-surface area of sand is small (Sutinen, 1992),its influence on electromagnetic waves is limited. Capillary forces result from pressure differ-ences between water and air phases in the pores and control the amount of free water in the porespace. The capillary pressure is inversely proportional to the pore radius (r). Matric forces thusexert suction on the pore water, usually expressed by the pressure head (h) or suction potential(pF), where pF = log |h|. The term ’field capacity’ is defined as the specific suction potentialwhen drainage under the influence of gravity has ceased. The natural unsaturated conditionsreferred to in this paper are field-capacity conditions at pF 2.

The relation between suction potential and volumetric water content (θ) is a fundamentalsoil hydrological property (Arya and Paris, 1981) that is related to the size distribution andconnectivity of the pores (Vogel and Roth, 1998; Fens, 2000). Pore bodies with wide entrychannels (pore throats) will drain at low suction, whereas those with narrow channels drain athigher suction (Bouma, 1977). It is generally accepted that, at field-capacity conditions andfor equal porosity values, fine-grained sediment with small pores has a higher θ than coarse-grained sediment. The same holds for poorly sorted sediment relative to well-sorted sediment.Simple laboratory experiments allow measurement of the θ-pF relation. In the experiments,initially saturated sediment cores are drained and water content is measured at specific suctionpotentials. Water-retention curves can be modeled by fitting the measured θ-pF pairs with VanGenuchten (1980):

θ = θr +θs −θr

(1+(αvg|h|)n)mvg. (5.1)

Here, θr and θs represent the residual and saturated water contents, respectively, αvg [cm−1]and n are parameters to fit the shape and position of the curves, and mvg = 1 − n−1. This

72


experimental method is time consuming and cannot account for small-scale texture variationsrelated to sedimentary structures that are present in the sample volumes.

Formulas that describe the relation between pore-size distribution and capillary forces alsoallow calculation of the θ-pF relation. However, the pore-size distribution is difficult to mea-sure and, consequently, methods like network modeling (Vogel, 2000) and percolation theoryare based on simplified assumptions for pore structure. The good correlation between pore-sizedistribution and grain-size distribution (GSD) of eolian sediment (Van Den Berg et al., in re-view), allows the use of GSD as a proxy for water retention in non-saturated sediment (Mishraet al., 1989). Next to GSD, total porosity (η) is a factor of importance for the water retention insediment as it is generally a measure for the packing. Assuming sediment with equal grain-sizedistribution, tighter packing results in smaller pores that hold capillary water with greater ease.Neural network models allow calculation of the θ-pF relation of sediment using its GSD and η(Schaap and Bouten, 1996).

5.3 Test site

5.3.1 Experimental procedure

The 11 × 16-m test site was surveyed with a grid of 2-D GPR lines (Figure 5.1), using apulseEKKO radar system with 25-, 100-, 225-, 450-, and 900-MHz antennas. 450- and 900-MHz antennas were used to image a small, 3 × 4-m area along the long axis of the grid ingreater detail (Table 5.1). Subsequent to the GPR measurements, 2 trenches (A and I), each3-m deep, were dug for detailed study and sampling of the sediment. Lacquer peels were madefor macroscopic study of textural characteristics. A few core plugs were collected for laboratorymeasurements of water retention (Figure 5.2). The core plugs with a volume of 45 cm3 were

Table 5.1: Details of GPR sections and common-mid-point measurements (Locations in Figure 5.1).

Name X0 , Y0 Xmax , Ymax f Step size[m] [m] [MHz] [m]

1 0.00 , 0.00 15.75 , 0.00 450 0.0501 6.00 , 0.00 10.00 , 0.00 900 0.0255 6.00 , 0.75 10.00 , 0.75 900 0.0256 0.00 , 1.50 16.00 , 1.50 450 0.0506 6.00 , 1.50 10.00 , 1.50 900 0.0257 6.00 , 2.25 10.00 , 2.25 900 0.0252 0.00 , 3.00 15.90 , 3.00 450 0.0502 6.00 , 3.00 10.00 , 3.00 900 0.025G 6.00 , 0.00 6.00 , 3.00 450 0.050G 6.00 , 0.00 6.00 , 3.00 900 0.025C 8.00 , 0.00 8.00 , 10.95 450 0.050C 8.00 , 0.00 8.00 , 3.00 900 0.025H 10.00 , 0.00 10.00 , 3.00 450 0.050H 10.00 , 0.00 10.00 , 3.00 900 0.025CMP 1 6.00 , 0.00 - 450 0.050CMP 4 8.00 , 1.50 - 900 0.020f = frequency.

73

Chapter 5

I23

truncationlevel

Unit 7Unit 6

Unit 4Unit 5

Unit 3Unit 2

Unit 1

horizontallamination

horizontal andparallel lamination

low-angle crossstratification

high-angle crossstratification

0

1

2

3

0 4 6 12 142 108

ESE WNW(a)Position [m]

0.2

0.4

0.6

0.8

7.79 8.868.357.246.740

I21I22

I23

1223.004 1223.003 1223.002 1223.001

1223.005

1223.006

1223.0071223.008

1222.008

Position [m](b)

I21pI22p

Legend:

Grain-sizesamples

1223.006

TDR section

Core plug

I21

Box core forthin sections

I22p

Dep

th [m

]D

epth

[m]

Figure 5.2: (a) Trench wall panorama showing sedimentary units and (b) lacquer peels of the trench. Thelacquer peels show faint foreset lamination indicated by thin dashed lines and the locations of grain-sizesamples (vertical spacing 0.05 m), TDR measurements, box cores, and core plugs.

sampled by pushing a metal ring into the sediment. The metal rings have walls of minimumthickness, so as to reduce sediment disturbance. Along vertical sections in the wall of trench I,TDR was used to obtain detailed information on dielectric properties in the subsurface, andsamples were collected for macroscopic analysis of textural properties (Table 5.2). Althoughthe TDR and GPR measurements are separated by a one-week time interval we expect no largeambiguities as we removed the wind-dried surface of the trench before measuring TDR data.

To obtain detailed information on the variation in textural characteristics and electromag-netic properties normal to the high-angle sedimentary bedding, we selected three sites I21, I22,and I23 in the detailed study area in trench I (Figure 5.2b). Both I21 and I22 were set at thelocation of known fine-grained foreset tops; I23 was set at a position where no foreset was vis-ible. To measure variation in electromagnetic wave velocity perpendicular to the bedding wemeasured 4 detailed TDR sections with a 0.02-m spacing at these locations (Figure 5.2b).

In an experimental approach to measure small-scale variations in textural characteristics andelectromagnetic properties, we collected three 0.05 × 0.08 × 0.15-m (height × width × length)undisturbed box cores at sites I21, I22, and I23 (Figure 5.2b). The box cores have thin metal

74


6 7 8 9 10Position [m]

0

0.5

1

Dep

th [m

]

0

10

20900 MHz

3 4

(b)

0

1

2

3

Dep

th [m

]

0

20

40450 MHz

1

2

0 2 4 6 8 10 12 14

Position [m]ESE WNW(a)

Figure 5.3: GPR sections from line 1 that allow correlation with the trench wall panoramas and lacquerpeels in Figure 5.2. The 450-MHz image (a) has the same horizontal and vertical scale as Figure 5.2a.In the 900-MHz image (b), which has a horizontal and vertical scale half that of Figure 5.2b, the frameshows the outline of the lacquer peels in Figure 5.2b. The numbered labels in the radar sections refer tothe text and represent: ©1 , reflection from the top of unit 6; ©2 , reflection from unit 2 and 4; ©3 , reflectionfrom foreset associated with sample I21 in Figure 5.2b; ©4 , reflection from foreset associated with sampleI22 in Figure 5.2b. Both radar sections are plotted with an AGC gain.

walls, so as to reduce sediment disturbance. In the laboratory, the box cores were impregnatedwith a blue epoxy-resin dye and thin sections were prepared perpendicular to the sedimentarybedding to reveal the maximum textural variation. Next, the thin sections were used for imageanalysis of the sediment characteristics and reconstruction of water-retention curves. Directcorrelation between the box-core results and TDR measurements is difficult because the box

Table 5.2: Textural characteristics of the samples from unit 7 in trench I. Clay and silt are defined as theweight fractions smaller than 2 µm and between 2 and 62 µm, respectively. The two most relevant coreplugs are shown in Figure 5.2.

N Grain size [µm] Clay content Silt content Organic matterMean Sd % Sd % Sd % Sd

Vertical transect 15 280.13 25.73 0.43 0.05 0.61 0.11 0.46 0.21Core plugs 19 268.79 12.08 0.57 0.05 1.07 0.13 0.10 0.16N = number of samples; Sd = standard deviation.

75

Chapter 5

cores were collected a few centimeters deeper into the quarry wall to avoid sediment disturbancecaused by the TDR pins. Also, before impregnation with epoxy resin about one-third of the boxcores was used to sample core plugs for measurement of water retention (Figure 5.2b).

5.3.2 Sedimentology and stratigraphy

The sediment at the test site is subdivided into 7 units. Units 1, 3, 5, and 7 consist of cleanwindblown sands, whereas units 2, 4, and 6 are soils (Figure 5.2a). The ages of sampled organicmaterial show that unit 6 was formed between 1260 and 1425 AD (Chapter 4, Table 4.3). Thesediment below unit 6 was deposited as the relatively low Older Dunes that formed on top ofa prograding coastal barrier system (Jelgersma et al., 1970). After about 1200 AD, a period ofcoastal erosion marked the onset of the formation of the so-called Younger Dunes. Increasedsediment supply led to the development of dunes that were up to 35-m high.

Unit 7 in trench I, on which the present study focuses, is characterized by high-angle cross-stratification, dipping at a maximum angle of 40◦ towards the southeast. We attribute the ab-sence of a sharp crest to a relatively rapid filling of the depression in which unit 6 was formed.The lacquer peels show two 0.2 to 0.3-m thick convex-upward foresets in unit 7 (Figure 5.2b).They are characterized by a top layer (around 5-cm thick) with slightly finer grain size than thebulk sediment. Also, sub-centimeter-scale lamination is present within the high-angle foresets.The photographs of the lacquer peels do not have enough detail and contrast to show this lam-ination. The average grain size of the studied sediment in unit 7 is around 280 µm (mediumsand). The average silt and clay content is 0.6% and 0.4%, respectively (Table 5.2).

5.3.3 GPR measurements

Fundamental properties that control the behavior of GPR signals are dielectric permittivity (ε),electrical conductivity (σ), and magnetic permeability (µ), which together define electromag-netic wave impedance (Z). Impedance contrasts in the subsurface cause part of the propagatingelectromagnetic energy, proportional to the magnitude of change, to be reflected. For most nat-ural sediments, variations in µ are insignificant (Daniels et al., 1988; Chapter 3). In low-lossmaterial such as clean, dry sand, the influence of σ on the electromagnetic signal is negligible atGPR frequencies (Davis and Annan, 1989). In contrast, ε plays an important role in both prop-agation and reflection of electromagnetic waves. If one defines relative permittivity εr = ε/ε0,where ε0 is the permittivity of free space, εr of water is around 80, whereas air and quartz havevalues of 1 and around 4, respectively. Thus, water content governs the relative permittivity andelectromagnetic wave impedance of sediment. The electromagnetic wave velocity (v) is foundby v = c0/

√εr, where c0 is the velocity in vacuum 3 108 m s−1). Radar-wave reflectivity can

be found from the difference in electromagnetic wave impedance at layer transitions and canbe simplified with RC = (

√εr2 −√

εr1)/(√

εr2 +√

εr1), where εr1 and εr2 are relative permit-tivity above and below the transition, respectively. The quality and sharpness of the reflectiondepends on the transition-zone width (vertical range over which εr changes), relative to thepredominant wavelength (W/λ). A sharp transition in εr with depth gives a sharp reflection.In contrast, a gradual change in εr causes signal dispersion. Annan et al. (1991) estimate thatW/λ < 0.3 gives a sharp reflection. Under natural unsaturated conditions the transition-zonewidth depends on the total difference in water content above and below the contrast as well ason the water-retention characteristics of the sediment above the contrast (Young and Sun, 1999).

76


4

3

250 300

1223.001

3

4

4 4.5

0

0.2

0.6

0.4

1223.002

3

4 4.5

0

0.2

0.6

(a) (b) (c)

0.4

Dep

th [m

]

Dep

th [m

]Relative permittivity Relative permittivity Grain size [μm]

Figure 5.4: Variation in relative permittivity (εr) for TDR sections (a) 1223.002 and (b) 1223.001 withdepth and (c) variation in grain size with depth (see Figure 5.2 for locations). The numbered labels ©3 and©4 correspond approximately to the clinoforms indicated in Figure 5.3. The fact that event ©3 in TDRsection 1223.002 and event ©4 in TDR section 1223.001 have about the same depth is the result of thedipping of the foresets. In the grain-size section, the fine grain sizes associated with the foreset tops inFigure 5.2 are superimposed on the large-scale trend in grain size.

Using common-mid-point measurements (Figure 5.1) the velocity for unit 7 was estimatedaround 0.12 m ns−1, which allowed for an accurate time-to-depth conversion. The 450-MHzGPR profile shows several continuous sub-horizontal reflections (©1 and ©2 , Figure 5.3a), whichresult from the soil horizons (Figure 5.2). Between meters 6 and 15 the upper 0.75 m showsdipping reflections. These reflections result from the high-angle cross-stratified dunes in unit 7(Figure 5.2). The 900-MHz GPR profile images several high-angle foresets in greater detail(©3 and ©4 , Figure 5.3b). Next to the reflections of the high-angle stratification, the 900-MHzimage shows some short sub-horizontal reflections (right of ©3 in Figure 5.3b). These reflectionsare low amplitude and cannot be linked to sedimentary structures in the lacquer peels. Theypossibly originate in a feature at the surface positioned around meter 6.6. The reflection fromthe top of the soil horizon below unit 7 is rather vague. This is partly caused by the large widthof the transition zone, W . For a transition zone of 0.1 m, as observed at the boundary betweenunit 7 and unit 6 (Chapter 4, Figure 4.4), and a velocity of 0.12 m ns−1, W/λ is 0.75 for the900-MHz signal. Consequently, the signal experiences dispersion and a decrease in reflectionamplitude.

77

Chapter 5

1223.008

4 4.53.5


0

0.08

0.04

0.12

1223.007

4 4.53.5


Dep

th [m

]

0

0.08

0.04

0.12

1223.005

Dep

th [m

]

1223.006

33

4

Figure 5.5: Four cross plots illustrating the influence of sedimentary structures on relative permittivity.TDR sections 1223.005, 1223.006, and 1223.007 were measured at locations of known foreset tops;section 1223.008 was taken at a position where no foreset was visible. See Figure 5.2 for the locations.The numbered labels ©3 and ©4 correspond to those in Figures 5.3 and 5.4.

5.4 TDR measurements

Time-domain reflectometry was developed to characterize the conductivity and water contentof soils through measurement of their electromagnetic properties. The method is based onthe propagation of an electromagnetic signal along a probe that is inserted into the sediment(Heimovaara and Bouten, 1990; Figure 2.2). Next to measurements of temporal changes, onecan use TDR to construct vertical profiles of electromagnetic properties (Topp and Davis, 1985;Van Dam and Schlager, 2000). The relative permittivity (εr) can be calculated from the traveltime of the signal in the sediment. The volumetric water content (θ) can be found by substitutionof εr in Equation (2.8).

Vertical TDR sections show that the relative permittivity (εr) in unit 7 varies between 3.75and 4.75 (Figure 5.4a,b), which is around the typical value of 4 for dry sand. The grain sizeis relatively small at the top and bottom of the measured section, and larger in the middle part(Figure 5.4c). Smaller grain sizes lead to a higher amount of capillary water to be present inthe sediment. The two finer-grained foreset tops that are present in the lacquer peels can beidentified on the TDR sections by a local increase in εr and are marked by labels ©3 and ©4 (Fig-ure 5.4a,b). They correspond approximately with the numbered clinoforms in Figure 5.3 andforeset tops in Figure 5.2. Events ©3 and ©4 are two of several similar but unexplained permittiv-ity excursions in the TDR sections (Figure 5.4), that may be related to the sub-centimeter-scalelamination observed in the lacquer peels.

78


To obtain detailed information on the variation in electromagnetic properties perpendicularto the foresets, TDR measurements with a 0.02-m spacing were conducted at the three sitesI21, I22, and I23. Three of the detailed sections were set at locations of known foreset tops;the fourth section was taken at a position where no foreset was visible (Figure 5.2b). The TDRmeasurements at foreset locations show a small but distinct increase in εr from around 3.75to values between 4.25 and 4.75 (Figure 5.5). The increase marks the transition to the fine-grained foreset top. The finer grain sizes of the foreset tops cause an increase in capillary water.The lower end of the 5-cm thick foreset tops is marked by a decrease in εr. The transition-zone width (W ) of increase and decrease in εr is around 0.04 m. W/λ for a 900-MHz GPRwave would be around 0.3, which is just at the transition between a sharp radar reflection andsignificant dispersion (Annan et al., 1991). The TDR measurements were used to constructmodels of electromagnetic wave impedance that allow synthetic modeling of GPR reflectionsfor the foresets. Results of this will be discussed later, along with the impedance models of thethin sections.

5.5 Analysis of thin sections

5.5.1 Image analysis

The primary objective of image analysis of thin sections in this study is to estimate variationsin water content and electromagnetic properties from textural characteristics on a finer scalethan TDR can measure. Here, the thin section from box core I21 is described (Figure 5.2b). Aspreviously mentioned, the positional constraint of the thin section image relative to the foresettop is not very strong. Therefore, comparison of the thin-section measurements and TDR sec-tion 1223.005 is difficult. Four high-resolution (176 × 176 pixels per mm2) digital photographswere taken with some overlap (covering an area of approximately 65 × 15 mm) in a transectalong the long axis of the thin section. We corrected for small spectral inhomogeneities betweenphotographs using histogram matching in image processing software ERDAS IMAGINE. Also,we used the software for maximum-likelihood classification to obtain binary images of solidgrains and pores. To prevent edge effects on the margins, the images were cropped. To ob-tain unbiased measures for textural parameters from the binary images we separated touchinggrains using a cutting procedure by Van Den Berg et al. (in press). These processing stepswere performed on the thin section images to get the final image that was used for the actualmeasurements (Figure 5.6).

The final image was analyzed for a series of textural characteristics, including grain-sizedistribution (GSD) and porosity (η), using the method described in Van Den Berg et al. (inreview). With most textural variation present in the direction perpendicular to the sedimentarylayering we used a rectangular measurement window to estimate the textural characteristics inthe thin section (Figure 5.6). We assume textural variation parallel to the bedding (the longaxis of the measurement window) to be constant. The measurement window had a height of1.5 mm and was moved with increments half the height of the window to obtain 85 samples(Figure 5.6).

79

Chapter 5

1

10

20

30

40

50

60

70

80

855 mm

Figure 5.6: Digitized thin section of box core I21.The image is based on four high-resolution (176× 176 pixels per mm2) overlapping photographsthat were taken perpendicular to the sedimentarystructures under an angle with the vertical (Fig-ure 5.2b). Here, the image is tilted towards verti-cal such that the sedimentary layering appears ap-proximately horizontal. The ruler to the right ofthe image shows the positions of the 85 samples.For sample 1 the measurement window is shown.Marked samples 20, 27, 61, and 69 have charac-teristic texture (Figure 5.7a,b) and are analyzed inFigure 5.9.

80


200 250 300 0.35 0.40 0.04 0.05 0.06 3.5 4.0 4.5

Mean grain size [ m] Porosity [m3m-3] Volumetric water content[m3m-3]


10

20

30

40

50

60

70

80

Sam

ple

(a) (b) (c) (d)

Figure 5.7: Vertical transects of (a) mean grain size and (b) porosity along the long axis of thin sec-tion I21. (c) The water-content at pF 2 along the same transect was calculated using a neural networkmodel that describes the θ-pF relationship based on cumulative grain-size distribution and porosity. (d)The relative permittivity was calculated using a dielectric mixing model (Equation (5.3)). Open circlesindicate specific samples 20, 27, 61, and 69 that are analyzed in Figure 5.9.

The measured 2-D grain size was converted to 3 dimensions by applying a multiplicationfactor of 4/π, which is strictly spoken only valid for spheres. We accepted the assumption asa reasonable simplification. The mean grain size shows a variation between 200 and 325 µm(Figure 5.7a), which is comparable to the variation that was found using laser particle-size mea-surements (Figure 5.4c). The variation in mean grain size shows a coarsening-downward trendfrom sample 85 to 20 and a subsequent and sudden drop in mean grain size below sample 20.The porosity (η) is easily found from the binary image and varies between 0.32 and 0.4 (Fig-ure 5.7b). Around sample 63 there is a distinct step in η whereas mean grain size varies little.There is also a large step in porosity between samples 20 and 25. Although the zone with a largevariation in porosity coincides with a large variation in mean grain size (samples 20-45) there is

81

Chapter 5

0.3

0.35

0.4

200 250 300

Mean grain size [μm]

Por

osity

[m3 m

-3]

Figure 5.8: Cross plot of porosity versus mean grain size for thin section I21.

no correlation between mean grain size and porosity (Figure 5.8). The transitions in grain sizeand porosity (Figure 5.7a,b) are not always as sharp as they appear in the image (Figure 5.6).This probably is the result of the moving average window and the fact that transitions may notbe perfectly perpendicular to the long axis of the image.

5.5.2 Estimated small-scale water-retention characteristics

To estimate the θ-pF relations from the textural characteristics of the thin section we used aneural network model that was developed for similar sediment as in the present study and re-quired input of cumulative grain-size distribution and porosity. The original data set comprised204 undisturbed core plugs of 15 sandy forest soils in The Netherlands. The samples were takenbetween 0 and 1-m depth and contain little clay and organic matter (Schaap and Bouten, 1996).Neural networks are black-box models that are able to learn relations between data without priormodel concept (Schaap, 1996), which makes the method very useful for analyzing data sets thatlack understanding of all physical relations.

The 85 GSD’s that were measured from the thin section were fitted with Haverkamp andParlange (1986):

GSD(d) =1(

1+(dgsdd−1)ngsd)p . (5.2)

Here, GSD(d) is the cumulative weight fraction, d is grain size, dgsd and ngsd are measures of theaverage grain size and grain-size uniformity, respectively, and p = 1−n−1

gsd . The average root-mean-square error of the measured and fitted cumulative grain-size distributions was 0.019.For 4 specific measurements with contrasting mean grain size and porosity (Figure 5.7a,b),the fitted grain-size distributions (Figure 5.9a) illustrate the variation in grain size and sorting.Samples 61 and 69 are fine grained and have a narrow GSD (well sorted). Samples 20 and 27

82


are coarser grained, with sample 27 being relatively well sorted and sample 20 poorly sorted(Table 5.3). Also, the samples show a difference in porosity (Figure 5.7b). Samples 27 and 61have a low porosity whereas 20 and 69 have a high porosity (Table 5.3). Although the porosityis somewhat underestimated (Table 5.3) the fitted GSD’s agree well with the measured GSD ofcore plug I21p (Figure 5.9a).

Next, the neural network model M6 of Schaap and Bouten (1996) was used to constructwater-retention curves for each of the 85 measurements (Table 5.4). Giving input values fordgsd , ngsd , gravel percentage, organic-matter percentage, and porosity, the neural network gen-erated output values for the Van Genuchten (1980) parameters αvg and n in Equation (5.1).Curves for the 4 specific measurements illustrate the interplay between porosity and grain-sizedistribution in the control of water retention (Figure 5.9b). The porosity determines the sat-urated water content at zero suction, whereas the grain-size distribution and porosity togetherdetermine the inflection point and the steepness and height of the plateau in the curve. Samples61 and 69 have similar grain sizes but the higher porosity of 69 causes a higher saturated watercontent. The larger mean pore size (higher porosity) is due to looser packing and leads to asteeper plateau and an inflection point at a lower suction potential. Thus, the sediment at sam-ple 69 drains quicker and its water-retention curve crosses that of sample 61. Samples 20 and27 have comparable mean grain size but the sediment of sample 20 is poorly sorted and has thewidest distribution of grain sizes of all four samples (Figure 5.9a). This distribution leads toa water-retention curve that is gentler than the others (Figure 5.9b). Although the sediment forsample 20 starts to drain at the lowest suction values (between pF 0.5 and 1) it drains slowest(see value for n in Table 5.4).

Comparison of the four specific measurements with the core plug shows that although theplateaus do not exactly match, the overall pattern is very similar (Figure 5.9b). At pF 2, whichis around field-capacity conditions (Chapter 4, Table 4.7), the results for the core plug I21p andthe neural network results differ by a factor 3. The difference may be caused by the neuralnetwork modeling (the original data set in Schaap and Bouten (1996) contains more silt) or bymacro-porosity in the core plug. In modeling GPR reflections, relative contrasts in water contentand dielectric properties are more important than the absolute values. We therefore assume thedifference around pF 2 as a reasonable simplification. For all 85 samples, Figure 5.7c showsthe vertical variation in θ at pF 2, and ranges from around 0.04 to 0.06 (Table 5.5).

Table 5.3: Textural characteristics of the samples from unit 7 in trench I. Clay and silt are defined as theweight fractions smaller than 2 µm and between 2 and 62 µm, respectively. The two most relevant coreplugs are shown in Figure 5.2.

Sample N Grain size [µm] ηMean Sd [m3m−3]

20 321 303.325 144.373 0.37627 387 285.545 104.758 0.32561 615 209.996 64.805 0.33569 553 223.066 69.790 0.377Average for 1-85 492 245.852 89.752 0.358I21p - 293.209 85.879 0.446N = number of samples; Sd = standard deviation;η = porosity.

83

Chapter 5

(b)

0 0.2 0.40

2

4

20

27

61

69


Suc

tion

pote

ntia

l

2027

61

69I21p

(a)

20

27

61

69I21p

800 400 00

0.5

1

Grain size [μm]

Cun

ulat

ive

grai

n-si

ze d

istr

ibut

ion

Figure 5.9: Plots of (a) cumulative grain-size distribution (GSD) and (b) water-retention characteristicsfor samples 20, 27, 61, and 69 in thin section I21 and for core plug I21p. The 4 specific samples inthin section I21 are characteristic in mean grain size (Figure 5.7a) and porosity (Figure 5.7b). Thepositions are shown in Figure 5.6. The cumulative GSD’s from the thin section were calculated usingVan Den Berg et al. (in review) and fitted with Equation (5.2). The grain size from the core-plug sediment(location in Figure 5.2a) was measured with a laser-particle sizer. The θ-pF relation was calculated usinga neural network model (Schaap and Bouten, 1996) for the thin section samples and using standard labmeasurement techniques (section 4.2.3) for the core plug. The open circles represent the lab data forI21p; the solid line is the fit with Equation (5.1).

84


5.5.3 Dielectric properties

Under the reasonable assumption that the influence of adsorbed water can be ignored, a simplethree-phase dielectric mixing model was used to obtain relative permittivity from θ (Bohl andRoth, 1994):

εr = (θεαmw +(1−η)εαm

s +(η−θ)εαma )1/αm . (5.3)

Here, η is porosity, εw is relative permittivity of water, εs is relative permittivity of solid mate-rial, εa is relative permittivity of air, and αm is a constant (0.5 for isotropic and homogeneousmaterial; Roth et al., 1990). For εs we used a value of 3.7 (Chapter 3). The results for the 4specific samples are given in Table 5.5. Figure 5.7d shows that the εr values lie between 3.6and 4.5. Both absolute values and observed range are comparable to those found with the TDRmeasurements (Figure 5.4a,b). The most distinct step in permittivity lies between samples 60and 70 and is associated with variation in porosity. The major step in grain size at sample 20(Figure 5.7a) did not result in a prominent permittivity change.

5.6 GPR synthetic modeling

Using the average of the detailed TDR measurements 1223.005 and 1223.006 (Figure 5.5) andthe εr results for the thin section (Figure 5.7d), we constructed two models of electromagneticwave impedance. Figure 5.10 shows the reflection coefficient (RC) based on these impedancemodels versus depth. The TDR-based model has 0.02-m measurement increments whereas thedetailed thin-section model has 0.75-mm measurement intervals. To construct synthetic 1-DGPR traces from the impedance models we used pulseEKKO software (Sensors&Software,1996). The software transforms the impedance model from a depth scale into a time scale,followed by computation of the impulse response for the layered model. Next, all generatedreflections are calculated, including multiples and interlayer reflections. The ground responseis obtained by convolution of a standard pulseEKKO wavelet (both 450 and 900 MHz) with

Table 5.4: Neural network input and output for 4 specific samples in thin section I21 and for the averageof all 85 samples. Values for input parameters gravel percentage and organic-matter percentage wereset to zero whereas θs input was set equal to porosity (Table 5.3) and dry bulk density was calculatedthrough ρb = 2.65(1−η). Output parameters αvg and n define the position of the inflection point andthe amount of curvature in the θ-pF relationship (Equation (5.1)), respectively. The marginal differencebetween porosity (Table 5.3) and modeled saturated water content output can be neglected. Laboratorymeasurement results for core plug I21p are given for comparison (in italics).

Sample Model input Model outputdgsd [µm] ngsd ρb [g cm−3] αvg [cm−1] n θs [m3m−3]

20 346.856 4.080 1.654 0.032 2.614 0.37927 311.210 5.533 1.789 0.027 2.871 0.32661 224.326 6.841 1.762 0.019 3.512 0.33669 238.855 6.819 1.651 0.024 3.576 0.379Average for 1-85 266.531 5.529 1.700 0.025 3.153 0.360I21p - - 1.467 0.024 6.99 0.446dgsd = measure for average grain size; ngsd = GSD uniformity; ρb = dry bulk density;θs = saturated water content.

85

Chapter 5

the impulse response. One of the assumptions of the software is a vertically incident wave,which we accepted as a reasonable simplification. We ignored losses from spherical-waveformspreading because of the limited thickness of the model. Attenuation was set at a constant value(0.01 dB m−1) because of the limited variation in dielectric properties.

In the reflectivity model for the TDR data (Figure 5.10a), the first two events at 0.22 and0.24-m depth are associated with the increase in εr due to the transition to the fine-grainedforeset top (Figure 5.5). The next two negative events at 0.26 and 0.28-m depth are due to thesubsequent increase in grain size and decrease in εr at the bottom of the around 5-cm thickforeset top. Figure 5.11 shows the individual contribution of each RC event to the compositereflection (stack). In the 900-MHz synthetic trace the first two RC events cause clearly visi-ble excursions to the left (Figure 5.11a). This is an illustration of signal dispersion (indicatedby label ©1 in Figure 5.10a), and is due to the gradual transition in εr. The next two eventscause reflections with an opposite polarity and partly overlap the two previous reflections (Fig-ure 5.11a). Here, the overlap causes a large width of the positive excursion in the compositesignal between 0.55 and 1.2 ns, relative to the negative excursion between 0 and 0.55 ns. Thefinal two events are too small to recognize in the stacked signal. In the 450-MHz synthetic tracethe resolution is too low to distinguish any of the events separately. This image is a stack of6 different, overlapping waveforms, producing one composite reflection. Figure 5.11b clearlyillustrates that the 450-MHz wavelength is too long to discern both the top and bottom of thearound 5-cm thick foreset top (λ/4 ≈ 7.5 cm). Here, the first excursions of negative RC events3 and 4 coincide with the second excursions of events 1 and 2, and cause a composite reflectiondominated by constructive interference. Accurate comparison of these results with the originalGPR images is difficult because of the assumptions in the synthetic modeling. The modeledand original data are different for angle of wave incidence. Also, losses by attenuation and byspherical-waveform spreading are difficult to accurately model for the part overlying the foresettop. Thus, reflection strengths cannot be compared.

In the detailed thin-section model (Figure 5.10b), with data points 0.75 mm apart, which isa small fraction of the wavelength (Table 2.1), it is obvious that the vertical resolution of the

Table 5.5: Volumetric water content (θ) and relative permittivity (εr) at pF 2, for 4 specific samples inthin section I21, for the average of all 85 samples and for core plug I21p (in italics). For the thin sectionsamples θ was estimated using the neural network (Table 5.4). For the core plug the water-retentioncharacteristics were measured in the lab and fitted with Equation (5.1) to find θ at pF 2. Relative permit-tivity was calculated from Equation (5.3). The θ and εr values for the core plug are low compared to thethin-section samples. The steep plateaus in the water-retention curves partly cause the large differencebecause θ is very sensitive to small variations in the height of the plateau.

Sample θ [m3m−3] εr

20 0.056 4.0927 0.048 4.0261 0.059 4.3369 0.039 3.56Average for 1-85 0.052 4.03I21p 0.010 2.53θ = volumetric water content;εr = relative permittivity.

86


0.22

0.26

-0.03 0.03

3

2

RC - thin section

Dep

th [m

]

0.22

0.3

-0.03 0.03RC - TDR data

Dep

th [m

]

0.26

THI0059n

900 MHz

4

2

4

6

twt [

ns]

THI0054n

450 MHz

2

4

6

twt [

ns]

TDR0059n

900 MHz

1

2

4

6

twt [

ns]

twt [

ns]

TDR0054n

450 MHz

2

4

6

(b)

(a)

Figure 5.10: Radar reflectivity and synthetic radar traces constructed from (a) TDR measurements1223.005 and 1223.006 in Figure 5.5 and (b) the image analysis results in Figure 5.7d. The impedancemodels used for the synthetic modeling assume a 0.2-m thick homogeneous layer before the first event.The impedance and RC models have a depth scale whereas the synthetic traces have a time scale. Sincewave velocity varies with depth, relative thickness of the layers changes. The synthetic radar traces havethe same horizontal scale. The dashed wiggle shows the wavelet that was used to convolve with theimpulse response model. The numbered arrows ©1 - ©4 are referred to in the text.

87

Chapter 5

λ41 stack

-8000

0

8000

2

3

4

6

450 MHz

1 20 3

twt [ns]

Am

plitu

de

(b)

(a)

-5000

0

5000

stack

1

2

3

4

5

6

900 MHz

1 2

λ41

twt [ns]0

Am

plitu

de

5

Figure 5.11: Diagrams showing the stack of 6 interfering reflections from the TDR-based impedancemodel at (a) 900 MHz and (b) 450 MHz. The numbers 1 to 6 refer to the events in the TDR reflectivitymodel (Figure 5.10a). The composite trace is dominated by destructive and constructive interference. Weignored multiples and interlayer reflections as we consider them to contribute little to the total reflection.

GPR signal is far too low to resolve the thin lamination. Moreover, radar antennas transmit theirenergy in a complex three-dimensional pattern (Van Der Kruk, 2001), troubling the attempt tostudy the radar response of thin lamination using 1-D synthetic models. Nevertheless, fromseismic studies we know the importance to stay a few steps ahead of maximum resolution (e.g.,Mayer, 1979; Bracco Gartner, 2000) and to understand the behavior of individual wave trains;not only the behavior of the wave form propagating in three dimensions. As is seen in the 900-MHz synthetic trace, individual reflections cannot be distinguished, but groups of events can.The negative RC′s between 0.205 and 0.215 m (indicated by arrow ©2 in Figure 5.10b) together

88


form the excursion to the right in the 900-MHz trace. The group of positive RC′s between 0.21and 0.22 m (indicated by arrow ©3 in Figure 5.10b) cause the stepped pattern in the synthetictrace (indicated by arrow ©4 in Figure 5.10b). Below this the interference by the numerousoverlapping waveforms becomes too complicated to distinguish RC events in the synthetic im-age. The most important indication that the reflected signal is a composite of interfering signalsis illustrated by the difference in reflected signal for two different frequencies. The 450-MHzsynthetic produces a totally different trace than the 900-MHz one. Also, it becomes apparentfrom the synthetic models that the reflection pattern in the thin-section synthetics (Figure 5.10b)has lower mean amplitude than the reflection pattern from the TDR model (Figure 5.10a). Thisshows that interference from thinly layered sediment reduces the mean amplitude of radar re-flections. Although at the scale of the individual wave train the RC′s in Figure 5.10b reach thoseof the TDR-based model, a wavefront propagating in three dimensions will cause a decrease inthe amplitude of the interference pattern. In an actual measurement with a 3-D propagatingwavefront reflections from changes in permittivity that are thicker and laterally more contin-uous will overprint the low-amplitude interfering reflections. Kruse and Jol (submitted) showhow thin layers with small impedance contrasts get overprinted by reflections from larger con-trasts. However, reflections from sub-centimeter-scale layering may cause loss of energy andmay explain, especially in the case of constructive interference and sediment with little along-bedding variation, reflections that do not represent original bedding.


This study began with the premise that although GPR has proven a useful technique for theimaging of sediments, detailed and quantitative knowledge about radar-wave reflection fromsedimentary structures is lacking. It is not known in great detail which textural characteristicscontrol water retention and, thus, reflections. Also, the minimum magnitude of change for areflection is unclear and little is known about the contribution of structures smaller than λ/4 tothe total reflection.

It is shown that water content, and thus GPR reflections in unsaturated sediment, is con-trolled by the size distribution and connectivity of the pore network. The exact characteristicsof the pore network are difficult to estimate. Instead, grain-size distribution and total porositycan be used to estimate water-retention characteristics in sediment. Although other texture pa-rameters such as grain orientation may have an influence on water retention we consider thepossible effects to be of minor importance. At field-capacity conditions (around pF 2), whenGPR is commonly used to study sediment, fine-grained and tightly packed material retains morewater than coarse and loosely packed material. The minimum required variation in grain sizenecessary to produce reflections structures is small. Even the smallest change in water con-tent will cause a contrast in dielectric properties and, thus, reflection of electromagnetic energy.Quantifying the minimum magnitude of change necessary to produce a reflection that is visiblein GPR field measurements is less straightforward.

Figure 5.4c shows a maximum textural variation of ∼80 µm for unit 7. TDR measurementsin this unit show that high-angle GPR reflections of fine-grained foreset tops are caused by εr

variations between around 3.75 and 4.75. Impedance models based on the TDR measurementsacross an around 5-cm thick foreset top allowed synthetic modeling of 450- and 900-MHz GPRreflections. The results show that the reflected signal experiences dispersion due to the gradualtransition in dielectric properties. The thickness of the foreset top is just around maximum

89

Chapter 5

resolution (λ/4) and reflections from the top and bottom of the foreset top interfere with eachother.

The digitized thin-section image rendered detailed information about variation in texturenormal to the bedding in an experimental attempt to model the radar response across of sub-centimeter-scale sedimentary layering. A neural network and a dielectric mixing model es-timated εr at field-capacity conditions to vary between 3.6 and 4.5, which agrees well withthe values measured with TDR. Models of electromagnetic wave impedance and reflectivityconstructed from this 6-cm long transect allowed modeling the GPR response. the GPR syn-thetics show that interference of reflected signals leads to reflection patterns that change withfrequency. However, the model is too small-scale with respect to both vertical and horizontalresolution of the GPR signal to compare with actual GPR reflections of thinly bedded sediment.Nevertheless, it is important to stress that interfering reflections from laterally continuous thinbedding may cause reflection patterns that do not represent primary bedding. Future studies areneeded to improve understanding of thin-layer reflections.

90

6

SynthesisThe objective of this thesis was to improve quantitative understanding on the origin of ground-penetrating radar (GPR) reflections in sediment. I approached the objective by using a com-bination of sedimentology, geophysics and especially petrophysics (see Figure 1.2). Chapter 2was devoted to the characterization of electromagnetic properties in the subsurface in order tounderstand which sediment properties may influence the GPR signal. This was accomplishedby using time-domain reflectometry (TDR) to measure the vertical variation in the velocity ofelectromagnetic signals. In the later chapters, TDR was combined with more specific petro-physical techniques to quantify the effects of diagenetic iron oxide, organic matter in soils, andprimary sedimentary structures on the GPR signal. The results provide insight into differentreflection types (isolated and composites) and in reflections at different scales (from millimeterto decimeter scale variations in electromagnetic properties).

This synthesis aims (a) to broaden the context of two petrophysical methods that were usedin the thesis and (b) to generalize some of the findings that were presented. I first discuss sev-eral aspects of TDR and measurement and modeling of water retention; petrophysical methodswhich were prominently used in this thesis but are relatively new in GPR research (Section 6.1).Then, for two causes of radar reflections, secondary precipitated iron oxide and thin layering insedimentary structures, I explore how widely applicable the assumptions and conclusions are(Section 6.2).

6.1 Petrophysical techniques


TDR measures the velocity of electromagnetic signals, which is primarily dependent on watercontent. In this thesis the technique is used to estimate the variation in relative permittivity(as a proxy for the GPR contrast parameter, electromagnetic wave impedance) in the subsur-face. As relative permittivity is a frequency-dependent property it is important to understandthe frequency ranges of both TDR and GPR. The used TDR system has a frequency range of300 kHz to 3 GHz. Although the frequency range of GPR antennas is different, the results forboth techniques are comparable (Weiler et al., 1998; Huisman et al., 2001). This is caused bythe existence of a frequency window in the hundreds of megahertz range where electromagneticwave methods as GPR and TDR can operate under low-loss and non-dispersive conditions. The”GPR window” is bounded by signal dispersion on the low end and scattering losses on thehigh end (Powers, 1997); effects that are enhanced when the amount of clay or water in thesediment increases. For the moist-to-dry sandy material in this thesis these effects are minimal,which explains why GPR and TDR results commonly show a good correlation. Nevertheless, in

Chapter 6

this thesis most TDR measurements show a slight overestimation of the electromagnetic wavevelocity relative to the velocity from CMP measurements. For estimation of velocity contraststhis difference is negligible.

Calibration measurements illustrate the reproducibility of TDR data, both excluding andincluding removal of the probe between measurements. TDR lab measurements in Chapter 3(Table 3.3) were all repeated between 5 and 7 times without removing the probe from the sam-ple. Standard errors are small and range between 0.002 and 0.024. The small differences thatdo exist can be attributed to electronic equipment instability and to errors in the semi-automaticwaveform analysis (Figure 2.2c). Table 6.1 summarizes the results of TDR measurements ina test laboratory with sandy material that was comparable to the eolian sediment in this thesis(see Van Der Kruk, 2001, section 7.6). I conducted repeated measurements with a verticallyinserted TDR probe in a line at the surface and at shallow depths between 0.1 and 0.25 m. Rel-ative permittivity shows low values at the surface (∼3.2 to 3.4) and values around 4 below adepth of 0.1 m. Standard deviations for different measurements from the same layers are small,which shows the stability of TDR measurements. These results illustrate the suitability of TDRto measure relative differences and contrasts in electromagnetic properties

6.1.2 Water-retention characteristics

A recurring issue in this thesis was the important relation between sediment properties andwater content in unsaturated sediment. It was shown in the previous chapters that iron ox-ides, soil-organic matter, and small texture variations in sedimentary layering all influence thewater-retention characteristics of the sediment. These variations in water retention influencewave velocity and cause impedance contrasts that lead to radar reflections. In Chapter 4 it wasshown that for layers with different water-retention characteristics, changes in the hydrologicalsituation (suction potential) have important implications for reflectivity of these layers.

Throughout this thesis I used the generally accepted Van Genuchten (1980) parameters todescribe water-retention curves in sediment. These curves illustrate the decrease in water con-tent with increasing suction potential. For the laboratory samples (Chapters 4 and 5), I used theactual measurements to calibrate the Van Genuchten (1980) parameters and to define the water-retention curves. However, it appeared that at low suction values the curves for clean sand didnot perfectly match with the laboratory measurements of water-retention characteristics. Withonly two parameters describing the shape of the curve (αvg and n) and two parameters giving

Table 6.1: Relative permittivity values from TDR measurements in a test laboratory including removalof the TDR probe between the measurements. The material has a mean grain size of 312.2 µm withstandard deviation of 16.2 µm. Measurements were conducted in a line along the surface and in a seriesof 0.4-m long lines at different depths. The pins were inserted vertically.

Pin length N Y0 Ymax X range εr Sd[m] [m] [m] [m]

Z980106.003 / 004 0.05 86 0.00 0.00 2.125 3.191 0.130Z980106.001 / 002 0.10 121 0.00 0.00 3.0 3.377 0.206Z980106.009 / 010 / 013 / 014 / 017 / 019 0.05 18 0.10 0.25 0.4 4.162 0.125Z980106.008 / 011 / 012 / 015 / 016 / 020 0.10 20 0.10 0.25 0.4 4.100 0.202N = number of samples; εr = relative permittivity; Sd = standard deviation.

92

Synthesis

positional constraints (saturated water content at pF 0 and residual water content at pF 4.2) itwas impossible to accurately calibrate for the gradual decrease in water content between sat-urated conditions and the sharp drop in water content that was observed between pF 1.5 andpF 2. In this thesis the main concern is at radar reflections in sediment at field-capacity condi-tions (around pF 2). I therefore considered the modeling errors at lower suctions as a reasonablesimplification.

For the thin section measurements (Chapter 5) I used a neural network to estimate the VanGenuchten (1980) parameters from physical sediment properties (grain-size distribution andporosity). The water-retention curves obtained using this approach are comparable to thosederived from the laboratory samples (Figure 5.9b). However, at field conditions around pF 2,which is the main area of interest, the thin section curves predict a higher water content thanthe laboratory curves do. However, assuming the relative differences among the thin-sectioncurves are correct, a small error in absolute values still allows to model radar-wave reflectionscorrectly.

There exists some debate on whether residual water content should be equal to zero or not.With increase of suction potential, free and capillary water gradually leaves the pore system.At higher suctions, adsorbed water is expelled from the sediment and water content slowlydecreases to zero. Therefore, θr should strictly spoken be set to zero. However at pF 4.2, thesuction potential that is commonly considered to represent dry conditions, not all the adsorbedwater is removed from the sediment (see Sections 4.2 and 5.2). Here, the residual water contentis larger than zero. The amount of adsorbed water still retained at pF 4.2 varies per sedimenttype and largely depends on the specific-surface area of the solid phase. In Chapter 5, thestudied material is all sand and the variation in specific-surface area and residual water contentis small. Because adsorbed water contributes little to the relative permittivity I assumed θr tobe zero in this chapter. In Chapter 4, the focus was on variation of dielectric properties withsuction potential. Because the specific-surface area of organic matter is markedly different fromthat of sand and because the textural variation is larger than in Chapter 5 I included variationsin θr in the model.

6.2 Causes of radar reflections - how widely applicable arethe conclusions?

6.2.1 Iron oxides

Iron is a common element in the earth’s system and hardly any rock is completely devoid of iron.Most sedimentary rocks contain iron-oxide minerals in varying nature and abundance (Cornelland Schwertmann, 1996). Different processes such as bacterial fermentation and burning playa role in the production of iron oxides and in the conversion of one type into other types of ironoxide (Maher, 1998). Sediment may contain both detrital and diagenetic iron oxide. An exampleof GPR-imaged detrital iron oxide is given by Jol et al. (1998) who linked the presence of amagnetite deposit in a coastal sequence to the reflection of GPR signals. Frequently, diageneticprecipitation of iron oxide in sediment concentrates around zones of fluctuating groundwaterlevels (e.g., Dekker et al., 1997). Iron oxide is reduced in the permanently wet, anaerobic zoneand moves upward where it oxidizes in the horizon that is alternately wet and dry.

93

Chapter 6

Sposito (1989) showed that goethite is the most abundant type of iron oxide in soils andsediment around the world. Sediment that contains goethite as the sole iron oxide is foundpredominantly in temperate, humid climates, but occurs ubiquitously in all climatic zones (Cor-nell and Schwertmann, 1996). Therefore, precipitated goethite that causes GPR reflections(Chapter 3) may well occur in many deposits around the globe. Goethite commonly occurs incombination with other oxyhydroxides and oxides of iron. The relative amount of each type ofiron oxide largely depends on climate. Oxyhydroxides like lepidocrocite and oxides like ferri-hydrite primarily occur in temperate and cool climates, whereas iron oxides such as hematite,maghemite, and magnetite are more abundant in warm climates (Cornell and Schwertmann,1996). Lepidocrocite, ferrihydrite, and hematite have electromagnetic properties that are com-parable to those of goethite (i.e., no significant magnetic component, Table 3.4). Also, theseiron oxides have a surface roughness comparable to goethite (e.g., Walker et al., 1978; Frank,1981) and will retain capillary water in sediment with greater ease than quartz will do. As aconsequence, different types of iron oxide will affect radar waves in a similar way as shown forgoethite in Chapter 3. However, in (sub) tropical climates where maghemite and magnetite maybe more abundant than goethite, the magnetic component of the radar signal will be affectedsignificantly (Table 3.4). For these iron oxides the effect of increased water retention due tosurface roughness will still be present but may be masked by the effect of increased relativemagnetic permeability. Future studies are necessary to understand GPR reflections for thesesituations.

Many iron oxides are highly mobile, easily dissolved and reprecipitated. Owing to theirdiagenetic character, oxide bands often occur at (high) angles to sedimentary bedding. Thus,as most sediment contains iron oxides, precipitation fronts that may generate ambiguous radarreflections are probably common around the globe.

6.2.2 Sedimentary structures

Depositional lamination is present in all sedimentary environments. The sediment studied in thisthesis is all of eolian origin and it is important to explore whether the results from Chapter 5can be extrapolated to laminated sediment in other depositional settings. Criteria to estimatethe similarity in radar response for laminated sediment of different depositional styles are layerthickness and textural variation. Interference is controlled by the wave length of the radarsignal and by layer thickness. Mean laminae thickness may vary considerably among differentdeposits and depends on aggradation rate and properties of formative bed waves (e.g., Bridge,1997). To fully understand the radar response from layered sediment it is necessary to considerboth the smallest-scale laminae and the large contrasts in impedance. In Chapter 5, the thin-section method allowed to estimate small-scale variations in dielectric properties, whereas theTDR measurements quantified the larger contrasts. This combination may well serve GPRmodeling of thin lamination in a wide range of sedimentary environments.

Many deposits have wider grain-size distributions than the eolian sediment in this thesis.The diversity in texture controls the variation in water retention and, hence, in impedance con-trasts. As a result, the range of reflection coefficients may be wider. Low-amplitude reflectionsfrom thin lamination will be overprinted by reflections from less abundant but more prominentcontrasts. For a thick-bedded and coarse-grained delta deposit (Smith and Jol, 1992), Kruseand Jol (submitted) show how, with increasing wavelength, thin layers with small impedancecontrasts lose expression relative to larger contrasts in the final GPR record. Similar responses

94

Synthesis

20 2515 twt [ns]

100

MH

z20

0 M

Hz

450

MH

z90

0 M

Hz

20 2515

twt [ns]

00.

05-0

.05

3.5

44.

51

1.2

1.4

1.6Depth [m]

rela

tive

perm

ittiv

ityR

C

(a)

(b)

(c)

(d)

(e)

(f)

Fig

ure

6.1:

Mod

els

of(a

)el

ectr

omag

netic

wav

eim

peda

nce

and

(b)

refle

ctio

nco

effic

ient

for

lam

inat

edse

dim

ent

base

don

the

mod

elin

Figu

re5.

9d.

The

impe

danc

em

odel

sus

edfo

rth

esy

nthe

ticm

odel

ing

assu

me

a1-

mth

ick

hom

ogen

eous

unit

(εr

=4)

befo

reth

efir

stm

easu

rem

ent,

and

sedi

men

tary

laye

rs1-

mm

thic

k.R

adar

trac

esw

ere

calc

ulat

edfo

ran

tenn

afr

eque

ncie

sof

(c)

100

MH

z,(d

)20

0M

Hz,

(e)

450

MH

z,an

d(f

)90

0M

Hz,

and

ata

sam

plin

gin

terv

alof

0.1

ns.

Tra

ces

are

plot

ted

with

cons

tant

mul

tiplie

rga

ins

of75

,50,

25,a

nd25

,res

pect

ivel

y.

95

Chapter 6

10

20

30

twt [

ns]

200 MHz 10

20

30

twt [

ns]

900 MHz

Figure 6.2: Synthetic radar sections at 200 MHz and 900 MHz generated by printing the synthetic tracesof the model in Figure 6.1 20 times side by side. The dissimilarity between the two sections indicatesthat they are largely produced by interference. Only the topmost reflection is stationary and correspondsto the initial change in impedance.

may be expected from other cyclic deposits. Examples for these are fluvial bars and tidal de-posits. In fluvial bars, reflections from channel-scale bounding surfaces between accretionaryunits (bedsets) will be more prominent than low-amplitude reflections from internal sedimen-tary structures (Bridge, 1993). However, when mud drapes are present within bedsets somecare needs to be taken in interpretation as they may produce high-amplitude reflections (Bristowet al., 2000b). In tidal sediments with alternating coarse (deposited by ebb and flood currents)and fine (deposited during flow reversals) layers, reflections from the coarse-fine transitions willoutweigh reflections from sedimentary structures within the layers.

The results of Chapter 5 demonstrated the effect of signal interference caused by thin sedi-mentary layering. The impedance model presented in Figure 5.7d spans only 6 cm and shouldbe extended to calculate the radar response of thicker laminated units and to model lower fre-quencies (larger wave lengths). Using the normally distributed εr in Figure 5.7d, I constructedan impedance model for a laminated unit 60 cm in thickness (Figure 6.1) with randomly gen-erated εr values. Following the same procedure for synthetic modeling of radar signals as inSection 5.6, but now including losses from spherical-waveform spreading, I constructed GPRtraces for antenna frequencies of 100, 200, 450, and 900 MHz (Figure 6.1). The dissimilaritiesamong the four sections again show that the reflections are the result of signal interference. 2-Dsections of the modeled traces demonstrate the effect that interference may have for the inter-preters eye (Figure 6.2), assuming no variation parallel to the bedding. In most high-energysedimentary environments this assumption is unlikely to be valid. However, in lacustrine andmarine deposits significant variation parallel to the bedding is often absent. Here, interference

96

Synthesis

Dep

th [m

] (v

= 0

.09

m n

s-1 )

0

2

4

6

8

Position [m]20 16 12 8 4 0

12 8 4

0

2

4

6

8

2

3

1

33

2 2 2

Position [m]12 8 4

Position [m]

0

2

4

6

8

1

3 33

22 22

(b)(a)

0

2

4

6

8

Dep

th [m

] (v

= 0

.09

m n

s-1 )

Position [m]20 16 12 8 4 0

Dep

th [m

] (v

= 0

.09

m n

s-1 )

Dep

th [m

] (v

= 0

.09

m n

s-1 )

Figure 6.3: Original (top) and interpreted (bottom) radar sections from Ossendrecht (based on Fig-ure 2.7). The diagram shows the sections for (a) 100-MHz and (b) 200-MHz antennae. The numberedlabels represent reflections that can be correlated between the two records: ©1 , reflection from soil withhigh impedance contrast; ©2 and ©3 , reflections from sedimentary structures.

such as shown in Figure 6.2 may dominate GPR records.

In spite of the above, it should be stressed again that in many cases the relatively low-amplitude signal from reflections of thin layering in sedimentary structures is overprinted bymore prominent reflectors such as iron-oxide bands and soils or by reflections from more promi-nent sedimentary structures such as the fine-grained foreset top in Chapter 5. Although small-scale variation in texture is more common than soils or iron oxide in sediment, reflections withhigher amplitudes commonly dominate GPR records. As an illustration I compare the 100- and200-MHz sections collected in Ossendrecht (Figure 6.3). The widths of the first arrivals (firsttwo excursions to the right that represent the direct ground and air wave) in both records showthe difference in wave length with frequency. In the upper 4 meters several of the reflectionscan be directly correlated between the two frequencies. One reflection is from a soil with a highimpedance contrast (indicated with ©1 in Figure 6.3). Other reflections that allow an almostone-to-one correlation between both frequencies probably represent sedimentary structures thathave enhanced dielectric values due to the presence of either windblown organic material orprecipitated iron oxide (©2 , dipping to the left, and ©3 , dipping to the right in Figure 6.3). A fewof the remaining reflections in the two sections show less correlation, which may be indicative

97

Chapter 6

of interference of radar waves. As a whole however, the overall character of the reflectionsremains comparable. The zone of continuous parallel reflections at 0 - 1.5 m, the zone of left-dipping reflections at 1.5 - 2.5 m, the zone of right-dipping reflections at 2.5 - 3.5 m that gradesinto the zone of discontinuous reflections at 3.5 - 5 m, and the zone of continuous sub-horizontalreflections at 5 - 7 m are visible in both records.

6.3 General conclusions

Identification of radar reflectors — TDR has proven to be very useful in understanding GPRreflections. It gives detailed insight into electromagnetic properties of sediment and, togetherwith sedimentological analyses and other petrophysical techniques, it allows identification andquantification of sediment characteristics that cause radar reflections.Causes of radar reflections — Radar reflections in sediment result primarily from variationsin water content that cause contrasts in dielectric properties and, thus, electromagnetic waveimpedance. Water-content variations in unsaturated sediment can be of different origin. Dif-ferences in water content can be caused by the presence of organic matter or of diagenetic ironoxide. Both organic matter and iron oxides prevent interstitial water from draining, which leadsto impedance contrasts with clean sand. Also, texture variation in sedimentary structures leadsto differences in water retention that can cause radar reflections.Character of radar reflections — Large impedance contrasts such as between soils and cleansand are strong reflectors. When transitions are sharp and when the layers are thick enoughrelative to the wave length of the radar signal, these transitions lead to prominent GPR reflec-tions. Thin lamination causes low-amplitude, interfering reflections. The reflection patterndiffers with frequency but is commonly overprinted by more prominent reflections. Diageneticfeatures such as bands of iron-oxide precipitates often cross-cut sedimentary bedding at highangles and may obscure radar reflections from primary sedimentary structures.

98

List of used symbols and acronyms

A Pulse amplitude [V]B Magnetic induction [Tesla]c0 Electromagnetic wave velocity in vacuum (3 108) [m s−1]d Grain size [µm]dgsd Measure of the average grain size in neural network [µm]f Frequency [Hz]g Gravitational acceleration (9.81) [m s−2]h Pressure head [m]H Magnetic field strength [A m−1]j

√−1Kp TDR probe constant [m−1]Lc Length of the TDR cable [m]Lp Length of the TDR probe [m]m Mass [kg]mvg 1−n−1 [-]M Magnetization [A m−1]n Parameter to fit θ-pF curve [-]ngsd Measure for uniformity of grain-size distribution [-]p 1−n−1

gsd [-]pF Suction potential (log |h|) [-]pw Water pressure [N m−2]r Pore radius [m]RC Reflection coefficient [-]R0 Resistance of cable tester and connectors [Ω]Rc TDR cable resistance [Ω m−1]Rtot Total resistance of sediment and TDR equipment [Ω]Rκ Susceptometer reading of magnetic susceptibility [-]t Time [s]Δts Travel time of electromagnetic pulse along TDR rods [s]v Propagation velocity of electromagnetic signal [m s−1]V Volume [m3]W Transition zone width (interval over which εr changes) [m]X Horizontal position [m]Y Vertical position [m]z Depth [m]Z Electromagnetic wave impedance [Ω]Zc TDR cable impedance [Ω]

Symbols

α Attenuation constant [dB m−1]αvg Parameter to fit θ-pF curve [cm−1]αm Constant in mixing model (0.5) [-]tanδ Loss tangent [-]ε Dielectric permittivity [F m−1]ε′′ Imaginary part of the dielectric permittivity [-]ε0 Dielectric permittivity of free space (8.854 10−12) [F m−1]εr Relative permittivity [-]

εa Relative permittivity of air (1) [-]εs Relative permittivity of solid material [-]εw Relative permittivity of water (80) [-]

η Porosity [m3m−3]θ Volumetric water content [m3m−3]

θr Residual water content [m3m−3]θs Saturated water content [m3m−3]

θg Gravimetric water content [m3kg−1]κ Volume-specific magnetic susceptibility [-]λ Wavelength [m]µ Magnetic permeability [H m−1]µ0 Magnetic permeability of free space (4π 10−7) [H m−1]µr Relative magnetic permeability [-]ρ Density [kg m−3]ρb Dry bulk density [kg m−3]ρw Water density [kg m−3]ρ∞ TDR reflection coefficient at long times [-]σ Electrical conductivity [S m−1]σDC Bulk electrical conductivity [S m−1]χ Mass-specific magnetic susceptibility [m3kg−1]ω Angular frequency (2π f ) [radians s−1]

AGC Automatic gain controlCMP Common mid pointEDX Energy-dispersive X-ray spectroscopyGPR Ground-penetrating radarGSD Grain-size distributionICP-AES Inductively coupled plasma atomic emission spectrometryPPM Parts per millionSEM Scanning electron microscopyTDR Time-domain reflectometryTGA Thermogravimetic analysis

100

References

Annan, A. P., Cosway, S. W. & Redman, J. D. (1991) Water table detection with ground penetratingradar. In: Society Exploration Geophysicists Annual Meeting, Houston, USA. Expanded Abstracts,pp. 494-496.

Arcone, S. A. (1984) Field observations of electromagnetic pulse propagation in dielectric slabs. Geo-physics, 49(10), 1763-1773.

Arya, L. M. & Paris, J. F. (1981) A physioempirical model to predict the soil moisture characteristicfrom particle-size distribution and bulk density data. Soil Science Society of America Journal, 45, 1023-1030.

Asprion, U. & Aigner, T. (1997) Aquifer architecture analysis using ground-penetrating radar: Triassicand Quaternary examples (S. Germany). Environmental Geology, 31, 66-75.

Barnhardt, W. A., Kayen, R. E. & Atwater, B. F. (1998) Radar-based stratigraphic records of the 1964Alaskan earthquake. In: GSA Annual Meeting, Toronto, Canada.

Bense, V. F., Van Den Berg, E. H. & Van Balen, R. T. (in review) Deformation mechanisms and hy-draulic properties of fault zones in unconsolidated sediments (Roer Valley Rift System, The Netherlands).Hydrogeology Journal.

Beres, M., Green, A., Huggenberger, P. & Horstmeyer, H. (1995) Mapping the architecture of glacio-fluvial sediments with three-dimensional georadar. Geology, 23(12), 1087-1090.

Berner, R. A. (1980) Early diagenesis: a theoretical approach. Princeton University Press, Princeton,241 pp.

Blazek, A. (1973) Thermal analysis. Van Nostrand Reinhold Company, London, 281 pp.

Bohl, H. & Roth, K. (1994) Evaluation of dielectric mixing models to describe the θ(ε)-relation. In:Time Domain Reflectometry in environmental, infrastructure and mining applications, Bureau of MinesSpecial Publication, 19, 309-317. US Department of Interior, Evanston.

Bouma, J. (1977) Soil survey and the study of water in the unsaturated soil - simplified theory and somecase studies. Netherlands Soil Survey Institute, Wageningen, 107 pp.

Bracco Gartner, G. L. (2000) High resolution impedance models of outcrops and their applications inseismic interpretation. Ph.D. thesis, Vrije Universiteit Amsterdam, 144 pp.

Bracco Gartner, G. L. & Schlager, W. (1999) Discriminating between onlap and lithologic interfinger-ing in seismic models of outcrops. AAPG Bulletin, 83(6), 952-971.

Brewster, M. L. & Annan, A. P. (1994) Ground-penetrating radar monitoring of a controlled DNAPLrelease: 200 MHz radar. Geophysics, 59(8), 1211-1221.

Bridge, J. S. (1993) The interaction between channel geometry, water flow, sediment transport anddeposition in braided rivers. In: Braided Rivers (Ed. by J. L. Best & C. S. Bristow), Geological SocietySpecial Publication, 75, 13-71. Geological Society, London.

Bridge, J. S. (1997) Thickness of sets of cross strata and planar strata as a function of formative bedwavegeometry and migration, and aggradation rate. Geology, 25, 971-974.

References

Bridge, J. S., Collier, R. E. L. & Alexander, J. (1998) Large-scale structure of Calamus River deposits(Nebraska, USA) revealed using ground-penetrating radar. Sedimentology, 45, 977-986.

Brisco, B., Pultz, T. J., Brown, R. J., Topp, G. C., Hares, M. A. & Zebchuck, W. D. (1992) Soilmoisture measurement using portable dielectric probes and time domain reflectometry. Water ResourcesResearch, 28(5), 1339-1346.

Bristow, C. S. (1995) Facies analysis in the Lower Greensand using ground-penetrating radar. Journalof the Geological Society, 152, 591-598.

Bristow, C. S., Pugh, J. & Goodall, T. (1996) Internal structure of aeolian dunes in Abu Dhabi deter-mined using ground-penetrating radar. Sedimentology, 43, 995-1003.

Bristow, C. S., Bailey, S. D. & Lancaster, N. (2000a) The sedimentary structure of linear sand dunes.Nature, 406(6), 56-59.

Bristow, C. S., Best, J. L. & Ashworth, P. J. (2000b) The use of GPR in developing a facies model fora large sandy braided river, Brahmaputra River, Bangladesh. In: 8th International Conference on GroundPenetrating Radar, University of Queensland, Australia. pp. 95-100.

Carcione, J. M. (1996) Ground-penetrating radar: wave theory and numerical simulation in lossyanisotropic media. Geophysics, 61(6), 1664-1677.

Carcione, J. M. (1998) Radiation patterns for 2-D GPR forward modeling. Geophysics, 63(2), 424-430.

Cardimona, S. J., Clement, W. P. & Kadinsky-Cade, K. (1998) Seismic reflection and ground-penetrating radar imaging of a shallow aquifer. Geophysics, 63(4), 1310-1317.

Casper, D. A. & Kung, K.-J. S. (1996) Simulation of ground-penetrating radar waves in a 2-D soilmodel. Geophysics, 61(4), 1034-1049.

Clement, W. P., Cardimona, S. J., Endres, A. L. & Kadinsky-Cade, K. (1997) Site characterizationat the Groundwater Remediation Field Laboratory. The Leading Edge, 16(11), 1617-1621.

Clemmensen, L. B., Andreasen, F., Nielsen, S. T. & Sten, E. (1996) The late Holocene coastal dune-field at Vejers, Denmark: characteristics, sand budget and depositional dynamics. Geomorphology, 17,79-98.

Cornell, R. M. & Schwertmann, U. (1996) The iron oxides - structure, properties, reactions, occurrenceand uses. VCH publishers, Weinheim, 573 pp.

Dalton, F. N., Herkelrath, W. N., Rawlins, D. S. & Rhoades, J. D. (1984) Time-domain reflectometry:Simultaneous measurement of soil water content and electrical conductivity with a single probe. Science,224, 989-990.

Daniels, D. J., Gunton, D. J. & Scott, H. E. (1988) Introduction to subsurface radar. IEE (Institution ofElectrical Engineers) Proceedings, 135(4), 278-320.

Davis, J. L. & Annan, A. P. (1989) Ground-penetrating radar for high resolution mapping of soil androck stratigraphy. Geophysical Prospecting, 37, 531-551.

De Marsily, G. (1986) Quantitative hydrogeology. Academic Press, London, 440 pp.

Dekker, L. W., Booij, A. H. & Ritsema, C. J. (1997) IJzerbanden en ijzerwanden in onze zanden.Stromingen, 3(2), 29-40.

Dekkers, M. J. (1990) Magnetic properties of natural goethite-III. Magnetic behaviour and propertiesof minerals originating from goethite dehydration during thermal demagnetization. Geophysical JournalInternational, 103, 233-250.

Doolittle, J. A. & Collins, M. E. (1995) Use of soil information to determine application of groundpenetrating radar. Journal of Applied Geophysics, 33(1-3), 101-108.

Driessen, P. M. & Dudal, M. (1991) The major soils of the world. Koninklijke Wohrmann B.V., Zut-phen, 310 pp.

102

References

Endres, A. L. (2000) Size scale considerations in modelling the electrical conductivity of porous rocksand soils. Exploration Geophysics, 31, 418-423.

Fagin, S. W. (1991) Seismic modeling of geologic structures - applications to exploration problems.Society Exploration Geophysics, Tulsa, 267 pp.

Fellner-Feldegg, H. (1969) The measurement of dielectrics in the time domain. The Journal of PhysicalChemistry, 73(3), 616-623.

Fens, T. W. (2000) Petrophysical properties from small rock samples using image analysis techniques.Ph.D. thesis, Delft University of Technology, 199 pp.

Fokkema, J. T., Fokkema, E., Beekman, S. & Slob, E. C. (2001) Analysis of georadar reflectionresponses. In: 7th International Congress of the Brazilian Geophysical Society, Salvador-Bahia.

Foldvari, M. (1991) Measurement of different water species in minerals by means of thermal derivatog-raphy. In: Thermal analysis in geosciences (Ed. by W. Smykatz-Kloss & S. S. J. Warne), Lecture Notesin Earth Sciences, 38, 84-101. Springer-Verlag, Berlin.

Frank, J. R. (1981) Dedolomitization in the Taum Sauk limestone (Upper Cambrian), Southeast Mis-souri. Journal of Sedimentary Petrology, 51(1), 7-18.

Fredlund, D. G. & Xing, A. (1994) Equations for the soil-water characteristic curve. Canadian Geotech-nical Journal, 31, 521-532.

Freeland, R. S., Yoder, R. E. & Ammons, J. T. (1998) Mapping shallow underground features thatinfluence site-specific agricultural production. Journal of Applied Geophysics, 40(1-3), 19-27.

Gawthorpe, R. L., Collier, R. E. L., Alexander, J., Leeder, M. R. & Bridge, J. S. (1993) Ground pene-trating radar: application to sandbody geometry and heterogeneity studies. In: Characterization of fluvialand aeolian reservoirs (Ed. by C. P. North & D. J. Prosser), Geological Society Special Publication, 73,421-432. Geological Society, London.

Giese, K. & Tieman, R. (1975) Determination of the complex permittivity from thin-sample time do-main reflectometry: improved analysis of the step response waveform. Advances in Molecular Relax-ation Processes, 7, 45-59.

Gochioco, L. M. (1992) Modeling studies of interference reflections in thin-layered media bounded bycoal seams. Geophysics, 57(9), 1209-1216.

Gupta, S. C. & Larson, W. E. (1979) Estimating soil water retention characteristics from particle sizedistribution, organic matter content, and bulk density. Water Resources Research, 15, 1633-1635.

Hanninen, P. (1992) Application of ground penetrating radar and radio wave moisture probe techniquesto peatland investigations. Geological Survey of Finland Bulletin, 361, 1-71.

Haverkamp, R. & Parlange, J. Y. (1986) Predicting the water-retention curve from particle-size distri-bution, 1, sandy soils without organic matter. Soil Science, 142, 325-339.

Heimovaara, T. J. (1993) Design of triple-wire time domain reflectometry probes in practice and theory.Soil Science Society of America Journal, 57, 1410-1417.

Heimovaara, T. J. & Bouten, W. (1990) A computer-controlled 36-channel time domain reflectometrysystem for monitoring soil water contents. Water Resources Research, 26(10), 2311-2316.

Heimovaara, T. J., Focke, A. G., Bouten, W. & Verstraten, J. M. (1995) Assessing temporal variationsin soil water composition with time domain reflectometry. Soil Science Society of America Journal,59(3), 689-698.

Heimovaara, T. J., De Winter, E. J. G., Van Loon, W. K. P. & Esveld, D. C. (1996) Frequency-dependent dielectric permittivity from 0 to 1 GHz: Time domain reflectometry measurements comparedwith frequency domain network analyzer measurements. Water Resources Research, 32(12), 3603-3610.

103

References

Hollender, F. & Tillard, S. (1998) Modeling ground-penetrating radar wave propagation and reflectionwith the Jonscher parameterization. Geophysics, 63(6), 1933-1942.

Hubbard, S. S., Rubin, Y. & Majer, E. (1997) Ground-penetrating-radar-assisted saturation and per-meability estimation in bimodal systems. Water Resources Research, 33(5), 971-990.Huggenberger, P. (1993) Radar facies: recognition of facies patterns and heterogeneities within Pleis-tocene Rhine gravels, NE Switzerland. In: Braided Rivers (Ed. by J. L. Best & C. S. Bristow), GeologicalSociety Special Publication, 75, 163-176. Geological Society, London.

Huggenberger, P., Meier, E. & Pugin, A. (1994) Ground-probing radar as a tool for heterogeneityestimation in gravel deposits: advances in data-processing and facies analysis. Journal of Applied Geo-physics, 31, 171-184.

Huisman, J. A. & Bouten, W. (2000) Comparison of calibration and direct measurement of cable andprobe properties in Time Domain Reflectometry. Soil Science Society of America Journal, 63(6), 1615-1617.

Huisman, J. A., Sperl, C., Bouten, W. & Verstraten, J. M. (2001) Soil water content measurementsat different scales: accuracy of time domain reflectometry and ground-penetrating radar. Journal ofHydrology, 245, 48-58.

Jackson, T. J. (1987) Effects of soil properties on microwave dielectric constants. US TransportationResearch Board, pp. 126-131.

Jelgersma, S., De Jong, J., Zagwijn, W. H. & Van Regteren Altena, J. F. (1970) The coastal dunes ofthe western Netherlands; geology, vegetational history and archeology. Mededelingen Rijks GeologischeDienst, Nieuwe Serie 21, 93-167.

Jol, H. M. & Smith, D. G. (1991) Ground penetrating radar of northern lacustrine deltas. CanadianJournal of Earth Sciences, 28, 1939-1947.

Jol, H. M., Peterson, C. D., Vanderburgh, S. & Phipps, J. (1998) GPR as a regional geomorphicmapping tool: shoreline accretion/erosion along the Columbia River littoral cell. In: 7th InternationalConference on Ground Penetrating Radar, University of Kansas, USA. pp. 257-262.

Kallweit, R. S. & Wood, L. C. (1982) The limits of resolution of zero-phase wavelets. Geophysics,47(7), 1035-1046.

Kasse, C. (1988) Early-Pleistocene tidal and fluviatile environments in the southern Netherlands andnorthern Belgium. Ph.D. thesis, Vrije Universiteit Amsterdam, 190 pp.

Klein, K. & Santamarina, J. C. (2000) Ferromagnetic inclusions in geomaterials: implications. Journalof Geotechnical and Geoenvironmental Engineering, 126(2), 167-179.

Knapp, R. W. (1990) Vertical resolution of thick beds, thin beds, and thin-bed cyclothems. Geophysics,55(9), 1183-1190.

Knoll, M. D. & Knight, R. (1994) Relationships between dielectric and hydrogeologic properties ofsand-clay mixtures. In: 5th International Conference on Ground Penetrating Radar, Waterloo Centre forGroundwater Research, Canada. pp. 45-61.

Konert, M. & Vandenberghe, J. (1997) Comparison of laser grain size analysis with pipette and sieveanalysis: a solution for the underestimation of the clay fraction. Sedimentology, 44, 523-535.

Kowalsky, M. B., Dietrich, P., Teutsch, G. & Rubin, Y. (2001) Forward modeling of ground-penetra-ting radar data using digitized outcrop images and multiple scenarios of water saturation. Water Re-sources Research, 37(6), 1615-1625.

Kruse, S. E. & Jol, H. M. (submitted) Amplitude and waveform analysis of repetitive GPR reflections:Lake Bonneville Delta, Utah. In: Ground penetrating radar in sediments: applications and interpretation(Ed. by C. S. Bristow & H. M. Jol), Geological Society Special Publication. Geological Society, London.

104

References

Maher, B. A. (1998) Magnetic properties of modern soils and Quaternary loessic paleosols: paleocli-matic implications. Palaeogeography Palaeoclimatology Palaeoecology, 137, 25-54.

Matzler, C. (1998) Microwave permittivity of dry sand. IEEE Transactions on Geoscience and RemoteSensing, 36(1), 317-319.

Mayer, L. A. (1979) Deep sea carbonates: acoustic, physical and stratigraphic properties. Journal ofSedimentary Petrology, 49(3), 819-836.

Mayer, L. A. (1980) Deep-sea carbonates: physical property relationships and the origin of high-frequency acoustic reflectors. Marine Geology, 38, 165-183.

Mishra, S., Parker, J. C. & Singhal, N. (1989) Estimation of soil hydraulic properties and their uncer-tainty from particle size distribution data. Journal of Hydrology, 108, 1-18.

Mullins, C. E. (1977) Magnetic susceptibility of the soil and its significance in soil science - a review.Journal of Soil Science, 28, 223-246.

Neal, A. & Roberts, C. L. (2000) Applications of ground-penetrating radar (GPR) to sedimentological,geomorphological and geoarchaeological studies in coastal environments. In: Coastal and estuarineenvironments: sedimentology, geomorphology and geoarchaeology (Ed. by K. Pye & J. R. L. Allen),Geological Society Special Publication, 175, 139-171. Geological Society, London.

Olhoeft, G. R. (1988) Selected bibliography on ground penetrating radar. In: Symposium on the Appli-cations of Geophysics to Engineering and Environmental Problems. Society of Exploration Geophysi-cists, Golden, CO, pp. 462-520.

Olhoeft, G. R. & Capron, D. E. (1994) Petrophysical causes of electromagnetic dispersion. In: 5thinternational conference on Ground-penetrating radar, pp. 145-152, Kitchener, Ontario, Canada.

Powers, M. H. (1997) Modeling frequency-dependent GPR. The Leading Edge, 16(11), 1657-1662.

Powers, M. H. & Olhoeft, G. R. (1995) GPRMODV2: one dimensional full waveform forward model-ing of dispersive ground penetrating radar data. US Geological Survey, Denver, 41 pp.

Robinson, D. A., Bell, J. P. & Batchelor, C. H. (1994) Influence of iron minerals on the determinationof soil water content using dielectric techniques. Journal of Hydrology, 161, 169-180.

Roksandic, M. M. (1995) Discussion on ’A 3-D seismic case history evaluating fluvially deposited thin-bed reservoirs in a gas-producing property’, by B. A. Hardage, R. A. Levy, V. Pendleton, J. Simmons, &R. Edsen with reply by the authors. Geophysics, 60(5), 1585-1592.

Roth, C. H., Malicki, M. A. & Plagge, R. (1992) Empirical evaluation of the relationship between soildielectric constant and volumetric water content as the basis for calibrating soil moisture measurementsby TDR. Journal of Soil Science, 43, 1-13.

Roth, K., Schulin, R., Fluhler, H. & Attinger, W. (1990) Calibration of time domain reflectometry forwater content measurement using a composite dielectric approach. Water Resources Research, 26(10),2267-2273.

Saarenketo, T. (1998) Electrical properties of water in clay and silty soils. Journal of Applied Geo-physics, 40, 73-88.

Saxton, K. E., Rawls, W. J., Romberger, J. S. & Papendick, R. I. (1986) Estimating generalized soilwater characteristics from texture. Soil Science Society of America Journal, 50(4), 1031-1036.

Schaap, M. G. (1996) The role of soil organic matter in the hydrology of forests on dry sandy soils.Ph.D. thesis, University of Amsterdam, 145 pp.

Schaap, M. G. & Bouten, W. (1996) Modeling water retention curves of sandy soils using neural net-works. Water Resources Research, 32(10), 3033-3040.

Schwan, J. (1991) Palaeowetness indicators in a Weichselian Late Glacial to Holocene aeolian succes-sion in the southwestern Netherlands. Zeitschrift fur Geomorphologie, N.F., Suppl.-Bd. 90, 155-169.

105

References

Schenk, C. J., Gautier, D. L., Olhoeft, G. R. & Lucius, J. E. (1993) Internal structure of an aeoliandune using ground-penetrating radar. In: Aeolian sediments: ancient and modern (Ed. by K.Pye &N. Lancaster), International Association of Sedimentology Special Publication, 16, 61-69. Blackwell,Oxford.

Sensors&Software (1996) pulseEKKO synthetic radargram user’s guide, Sensors&Software, Missis-sauga, 18 pp.

Smart, P. & Tovey, N. K. (1981) Electron microscopy of soils and sediments: examples. Oxford Uni-versity Press, 177 pp.

Smith, D. G. & Jol, H. M. (1992) Ground-penetrating radar investigation of a Lake Bonneville delta,Provo level, Brigham City, Utah. Geology, 20, 1083-1086.

Sposito, G. (1989) The chemistry of soils. Oxford University Press, New York, 304 pp.

Stafleu, J. & Sonnenfeld, M. D. (1994) Seismic models of a shelf-margin depositional sequence: upperSan Andres Formation, Last Chance Canyon, New Mexico. Journal of Sedimentary Research, 64(4),481-499.

Stafleu, J. & Schlager, W. (1995) Pseudo-unconformities in seismic models of large outcrops. Geolo-gische Rundschau, 84, 761-769.

Stuiver, M., Reimer, P. J., Bard, E., Beck, J. W., Burr, G. S., Hughen, K. A., Kromer, B., McCormac,G., Van Der Plicht, J. & Spurk, M. (1998) INTCAL98 radiocarbon age calibration, 24,000-0 cal BP.Radiocarbon, 40(3), 1041-1083.

Sutinen, R. (1992) Glacial deposits, their electrical properties and surveying by image interpretation andground penetrating radar. Geological Survey of Finland Bulletin, 359, 1-123.

Theimer, B. D., Nobes, D. C. & Warner, B. G. (1994) A study of the geoelectrical properties of peat-lands and their influence on ground-penetrating radar surveying. Geophysical Prospecting, 42(3), 179-209.

Thompson, R. & Oldfield, F. (1986) Environmental magnetism. Allen & Unwin, London, 227 pp.

Tinga, W. R., Voss, W. A. G. & Blossey, D. F. (1973) Generalized approach to multiphase dielectricmixture theory. Journal of Applied Physics, 44, 3897-3902.

Topp, G. C. & Davis, J. L. (1985) Measurement of soil water content using time-domain reflectometry(TDR): a field evaluation. Soil Science Society of America Journal, 49, 19-24.

Topp, G. C., Davis, J. L. & Annan, A. P. (1980) Electromagnetic determination of soil water content:measurements in coaxial transmission lines. Water Resources Research, 16(3), 574-582.

Topp, G. C., Davis, J. L. & Annan, A. P. (1982) Electromagnetic determination of soil water contentusing TDR: I. applications to wetting fronts and steep gradients. Soil Science Society of America Journal,46, 672-678.

Ulriksen, C. P. F. (1982) Application of impulse radar to civil engineering. Ph.D. thesis, Lund Universityof Technology, 175 pp.

Van Dam, R. L. (accepted) Internal structure and development of a Pleistocene eolian dune, The Nether-lands, from 3-D interpretation of ground-penetrating radar data. Netherlands Journal of Geosciences.

Van Dam, R. L. & Schlager, W. (2000) Identifying causes of ground-penetrating radar reflections usingtime-domain reflectometry and sedimentological analyses. Sedimentology, 47(2), 435-449.

Van Dam, R. L., Schlager, W., Dekkers, M. J., & Huisman, J. A. (accepted). Iron oxides as a causeof GPR reflections. Geophysics, scheduled for publication in the March/April 2002 issue, 67(2).

Van Dam, R. L., Van Den Berg, E. H., Van Heteren, S., Kasse, C., Kenter, J. A. M., & Groen, K.(accepted). Influence of organic matter in soils on radar-wave reflection: sedimentological implications.Journal of Sedimentary Research, scheduled for publication in the May 2002 issue, 72(3).

106

References

Van Dam, R. L., Van Den Berg, E. H., Schaap, M. G., Broekema, L. H., & Schlager, W. (submitted).Radar reflections from sedimentary structures in the vadose zone. In: Ground penetrating radar in sed-iments: applications and interpretation (Ed. by C. S. Bristow & H. M. Jol), Geological Society SpecialPublication. Geological Society, London.

Van Den Berg, E. H., Meesters, A. G. C. A., Kenter, J. A. M. & Schlager, W. (in press) Automatedseparation of touching grain sections in digital images. Computers and Geosciences, scheduled forpublication in the February 2002 issue, 28(1).

Van Den Berg, E. H., Bense, V. F. & Schlager, W. (in review) Textural characteristics and fabric inlaminated sand with digital image analysis of thin sections. Journal of Sedimentary Research.

Vandenberghe, J. & Van Overmeeren, R. A. (1999) Ground penetrating radar images of selectedfluvial deposits in the Netherlands. Sedimentary Geology, 128, 245-270.

Van Der Kruk, J. (2001) Three-dimensional imaging of multi-component ground penetrating radar data.Ph.D. thesis, Delft University of Technology, 242 pp.

Van Der Kruk, J., Slob, E. C. & Fokkema, J. T. (1999) Background of ground-penetrating radarmeasurements. Geologie en Mijnbouw, 77, 177-188.

Van Genuchten, M. T. (1980) A closed-form equation for predicting the hydraulic conductivity of un-saturated soils. Soil Science Society of America Journal, 44, 892-898.

Van Heteren, S. & Van de Plassche, O. (1997) Influence of relative sea-level change and tidal-inletdevelopment on barrier spit stratigraphy, Sandy Neck, Massachusetts. Journal of Sedimentary Research,67(2), 350-363.

Van Heteren, S., FitzGerald, D. M., McKinlay, P. A. & Buynevich, I. V. (1998) Radar facies ofparaglacial barrier systems: coastal New England, USA. Sedimentology, 45, 181-200.

Van Overmeeren, R. A. (1994) Georadar for hydrogeology. First Break, 12(8), 401-408.

Van Overmeeren, R. A. (1998) Radar facies of unconsolidated sediments in The Netherlands: A radarstratigraphy interpretation method for hydrogeology. Journal of Applied Geophysics, 40, 1-18.

Vogel, H. J. (2000) A numerical experiment on pore size, pore connectivity, water retention, permeabilityand solute transport using network models. European Journal of Soil Science, 51, 99-105.

Vogel, H. J. & Roth, K. (1998) A new approach for determining effective soil hydraulic functions.European Journal of Soil Science, 49, 547-556.

Von Hippel, A. R. (1954) Dielectric Materials and Applications. John Wiley and Sons, New York,438 pp.

Vorosmarty, C. J., Green, P., Salisbury, J. & Lammers, R. B. (2000) Global water resources: vulner-ability from climate change and population growth. Science, 289, 284-288.

Walker, T. R., Waugh, B. & Grone, A. J. (1978) Diagenesis in first-cycle desert alluvium of Cenozoicage, southwestern United States and northwestern Mexico. Geological Society of America Bulletin, 89,19-32.

Ward, R. C. & Robinson, M. (1990) Principles of Hydrology. McGraw-Hill Book Company, London,365 pp.

Warner, B. G., Nobes, D. C. & Theimer, B. D. (1990) An application of ground penetrating radarto peat stratigraphy of Ellice Swamp, southwestern Ontario. Canadian Journal of Earth Sciences, 27,932-938.

Waterloo, M. J. (1994) Water and nutrient dynamics of Pinus Caribaea plantation forests on formergrassland soils in southwest Viti Levu, Fiji. Ph.D. thesis, Vrije Universiteit Amsterdam, 478 pp.

Weerts, A. H., Bouten, W. & Verstraten, J. M. (1999) Simultaneous measurement of water retentionand electrical conductivity in soils: testing the Mualem-Friedman tortuosity model. Water ResourcesResearch, 35(6), 1781-1787.

107

References

Weiler, K. W., Steenhuis, T. S., Boll, J. & Kung, K.-J. S. (1998) Comparison of ground penetratingradar and time-domain reflectometry as soil water sensors. Soil Science Society of America Journal, 62,1237-1239.

Welton, J. E. (1984) SEM petrology atlas. The American Association of Petroleum Geologists, 237 pp.Wensink, W. A. (1993) Dielectric properties of wet soils in the frequency range 1-3000 MHz. Geophys-ical Prospecting, 41, 671-696.

Worm, H.-U., Clark, D. A. & Dekkers, M. J. (1993) Magnetic susceptibility of pyrrhotite: grain size,field and frequency depencence. Geophysical Journal International, 114, 127-137.

Worsfold, R. D., Parashar, S. K. & Perrott, T. (1986) Depth profiling of peat deposits with impulseradar. Canadian Geotechnical Journal, 23(2), 142-154.

Young, R. A. & Sun, J. (1999) Revealing stratigraphy in ground-penetrating radar data using domainfiltering. Geophysics, 64(2), 435-442.

Zegelin, S. J., White, I. & Jenkins, D. R. (1989) Improved field probes for soil water content and elec-trical conductivity measurement using time domain reflectometry. Water Resources Research, 25(11),2367-2376.

108

Acknowledgements

Completing this thesis would not have been possible without the help of many people, all ofwhich I want to thank greatly. There are a few people that I wish to mention in particular:

• Ronnie van Overmeeren is the one that introduced me to shallow geophysics during aninternship at TNO. His enthusiasm was very inspiring and contributed to my decision tostart this project. Without him making me aware of the vacancy I would not have had achance anyway.

• Sytze van Heteren has been of great importance during the whole project. As a GPRexpert and fieldwork enthusiast, he spent a lot of time with me in the field, which wasboth instructive and great fun. He was invaluable as discussion partner and as reviewerand co-author of several of my papers.

• Wolfgang Schlager, my promotor, has always been enthusiastic about the project andsuggested important topics that are covered in the thesis. With his broad knowledge heprovided reflection from many different angles.

• Elmer van den Berg, my roommate and hydrosed colleague, is the one I could not havemissed during the last 2 years of the research. His hydrological input was of enormousvalue to this thesis. We had a great deal of fun, but his helpfulness was sometimes to feelguilty about.

I wish to acknowledge the reading committee, consisting of Willem Bouten, Jacob Fokkema,Sytze van Heteren, and Sjef Meekes, for the quick judgement of the thesis and for the usefulsuggestions to improve it. I thank my co-authors Elmer van den Berg, Lucas Broekema, MarkDekkers, Koos Groen, Sytze van Heteren, Sander Huisman, Kees Kasse, Jeroen Kenter, MarcelSchaap, and Wolfgang Schlager for the contributions and discussions that helped building theseparate chapters of this thesis. I owe the reviewers of these chapters equally: Steve Arcone(3), Steve Cardimona (2), Ted Hickin (4), Peter Huggenberger (2), Ronnie van Overmeeren (5),Joseph Kruger (3), Brian Ricketts (4), and Joep Storms (2).

The TNO lunch meetings that were held four times a year were excellent opportunities to shareideas and sometimes find simple answers to difficult problems. I wish to thank all people thatregularly joined these GPR gatherings: Marcel Bakker, Wim van Dalfsen, Bart Goes, Marc vander Gulik, Vincent van Hoegaerden, Sander Huisman, Jan van der Kruk, Sjef Meekes, Sandervan Ouwerkerk, Ronnie van Overmeeren, and Rogier Westerhof.

No fieldwork no data, no data no thesis. I am therefore greatly indebted to all those peoplethat helped me out in Ossendrecht and Katwijk: the late Thom Roep, Jelmer Cleveringa, Kim

Acknowledgements

Cohen, Stefan Dekker, Michel Groen, Sytze van Heteren, Ko van Huisstede, Fabricio ’scalebar’de Jonge, Kees Kasse, Evert Slob, and Albrecht Weerts. Albert van Dijk and Jeroen Kenterwere of great importance as ’logistic managers’ in the Katwijk fieldwork.

Numerous people were instrumental in repairing equipment, making and digitizing thin sec-tions, solving computer problems, or doing lab measurements. Many thanks therefore go toFred Beekman, Arie Bikker, Gerard Kok, and Frans Stevens (computers), Johan de Lange andRon Lootens (electronics), Wim Bergenhenegouwen and Fred Schuurhof (photolab), SaskiaKars (SEM), Nanda Koot and Bouk Lacet (thin sections), and Marjan Reith (TGA).

I received a lot of help typesetting this thesis in LATEX from Patrck van den Boogaart, Richard deJeu, and especially Marco Kouwenhoven. Marco generously spoiled a few weekend days andskyrocketed his boss’ phone bill helping me with the layout. Talking of layout, I thank Gabrielade Aguiar for designing the cover.

Doing a Ph.D. is sometimes narrowing one’s view. I therefore am indebted to all those peoplewho gave me the opportunity to participate in research and fieldwork other than for my ownproject: Jef Vandenberghe, Jan de Bruin, and Sander Vlotman - GPR measurements in theNochten mine, Germany, 1997. Jelmer Cleveringa, Sytze van Heteren, Bart Schrijver, and Advan der Spek - GPR research in the coastal area of The Netherlands (Ypenburg, Vogelenzang,and Texel), 1997-1999. Jan van der Kruk - measurements in the TNO-FEL test lab, 1998. JacoBaas and Joep Storms - modeling the duration of turbidity currents, 1999. Victor Bense andRob Houtgast - GPR characterization of a fault plane in southeastern Netherlands, 1999. JohnBridge, Richard Collier, Ian Lunt, and Bo Tye - 3-D characterization of braided river deposits,Sagavanirktok River, Alaska, 2000.

At the university, Erwin Adams, Bernd Andeweg, Victor Bense, Remco van den Bos, Clau-dia Bouman, Boris van Breukelen, Hendrik Braaksma, Anne Fortuin, Guido Bracco Gartner,Heiko Hillgartner, Adrian & Ina Immenhauser, Bram van der Kooij, Klaudia Kuipers, VincentPost, Sam Purkis, Ute Satler, Jan Smit, John Woodside, Valentina Zampetti, Tiphaine Zitter,and many others provided a great work environment. I like to thank all participants in theMensa Posse, Geo Tourpool, VUCK (Vrije Universiteit Collegue Kartclub), and VUT (VrijeUniversiteit Tablefootball) for providing the necessary amount of distraction.

My special thanks are for my room mates Giovanna Della Porta, Ceciel van der Gulik, HandigeNico, and Florian ’potje?’ Maurer, who provided help, the right spirit, and humor when needed.

No work without a life; arriving home after a busy day, I was privileged to enter a flat with agreat atmosphere; thanks Arjen, Gabriela, Jurjen, Nathalie, Vicky, and Volker for everything.I owe all my friends (many of them in Utreg; others spread around the country or the world)for the great times when out of the VU. Thank you all! I would like to mention many, but inparticular Joep and Lucas, simply for being the best friends possible.

Finally, I wish to thank my family. My parents and sister have shown the one really importantaspect of life: ”To live and to enjoy it”. Thanks for all your support, interest, and love.

110

causes of ground-penetrating radar reﬂections in sedimentrvd/pub/rvddissertation.pdf · causes of...

Documents