· Title: Microsoft Word - 1.5.5_GPR.docx

British Society for Geomorphology Geomorphological Techniques, Part 1, Sec. 5.5 (2013)

1.5.5. Ground Penetrating Radar Martin Robinson1, Charlie Bristow2, Jennifer McKinley1 and Alastair Ruffell1 1School of Geography, Archaeology and Palaeoecology, Queen’s University Belfast 2Department of Earth and Planetary Sciences, Birkbeck University of London ([email protected])

ABSTRACT: Ground-penetrating radar (GPR) is an effective tool to visualise the structure of the shallow subsurface. The purpose of this article is to offer guidelines to non-specialist GPR users on the collection, processing and interpretation of GPR data in a range of environments. The discussion on survey design focuses on single fold, fixed-offset reflection profiling, the most common mode of GPR data collection, however the design factors can be applied to other survey types. Information on the visualisation of processed data, as well as the advantages and disadvantages of GPR, is provided. Possible applications of GPR in geomorphological research are presented, along with a case study outlining how GPR can be used to measure peat thickness.

KEYWORDS: Ground penetrating radar, survey design, processing, interpretation, applications

Introduction Ground-penetrating radar (GPR) has become a popular tool in sedimentological studies, with Table 1 illustrating recent research that has utilised GPR in the analysis of environmental processes (Jol and Bristow, 2003). It is a non-invasive geophysical technique designed primarily for subsurface investigation (Neal, 2004; Comas et al., 2004). A GPR system detects changes in the electrical properties of the shallow subsurface using discrete pulses of high frequency electromagnetic (EM) energy, usually in the 10-1000 MHz range (Neal, 2004). The technique has been successfully applied in a wide range of environmental studies however an understanding of the capabilities and limitations of GPR is vital when considering using the technique, with the quality of GPR results often being dependent on the surveyed environment (Daniels, 2004). The purpose of this article is to offer guidelines on the collection, processing and interpretation of GPR data. This paper presents an overview of good practice material for non-specialist users working in the field of environmental

research. An operator of GPR equipment must have an understanding of the fundamental principles underlying the technique (Daniels, 2004). A training course or field survey with an experienced operator is highly recommended.

Principles A GPR system transmits short pulses of high frequency EM energy (10-1000 MHz) from an antenna into the subsurface (Jol and Smith, 1991; Holden et al., 2002). As an EM wave disseminates downwards, its velocity is altered due to encounters with materials of differing electrical properties (Neal, 2004). Abrupt changes in the dielectric constant results in a portion of the energy being reflected, with the receiving antenna of the GPR system detecting the reflected EM energy (Figure 1a). Scattering of the radar waves occurs when the radar signal travels through overburden. It is worth noting that most GPR antennas are not focussed and transmit energy into the air as well as the ground. As a result it is possible to get reflections from objects above ground such as walls, cars, fences or overhead cables. Some antennas are shielded to reduce

ISSN 2047-0371

Ground Penetrating Radar 2


external noise and prevent the signals being transmitted through the air but this adds to the weight and bulk of the antennas which can be awkward in the field. The time between transmission and reception, referred to as the two-way travel-time (TWT) and commonly measured in nanoseconds, is a function of reflector depth and the EM velocity of propagation (Neal, 2004; Jol and Smith, 1991). GPR provides a continuous profile of the subsurface, displaying horizontal survey distance against vertical TWT. Vertical TWT is converted to depth with knowledge of the propagation velocity, expressed as (Equation 1):

𝑑 = 𝑣×𝑡/2 (Equation 1) In which 𝑑 is depth, 𝑣 is velocity, and 𝑡 is TWT. The propagation of EM energy through media is controlled by several material properties. Dielectric constant (dielectric permittivity), a property which is strongly dependent on the water content of a material, is the primary factor controlling the velocity of an EM wave. Reflections therefore can typically be related to interfaces where there is a considerable change in water content (Comas et al., 2005). Electrical conductivity is a measure of charge transport, through a

medium, on application of an electric field (Powers, 1997). The most important electrical conduction losses, in relation to GPR performance, occur due to ionic charge transport in water and electrochemical processes associated with cation exchange. Olhoeft (1998) describes the importance of clay mineral cation exchange in studies of soil and sediment. The equation for the velocity of propagation is expressed as (Equation 2):

𝑣 = 𝑐/ 𝜀! (Equation 2) In which 𝑣 is the velocity, 𝑐 is the speed of light (300mm/ns) and 𝜀! is the relative dielectric constant. Table 2 provides typical dielectric constant and electrical conductivity values for common materials encountered using GPR.

Suitability of GPR It is essential from a scientific view to clearly establish what data are required to test a particular hypothesis. This will influence the size of the survey, the depth of investigation and the resolution needed (Jol and Bristow, 2003). Before planning a survey it is important to determine if GPR will be effective.

Table 1. Recent environmental studies that use GPR

Depositional Setting Recent Papers

Glacial Appleby et al. (2010); Benediktsson et al. (2009); Degenhardt (2009); Gibbard et al. (2012); Hart et al. (2011); Irvine-Fynn et al. (2006); Kim et al. (2010); Langston et al. (2011); Leopold et al. (2011); Monnier et al. (2011); Murray and Booth (2010)

Fluvial Ashworth et al. (2011); Johnson and Carpenter (In Press); Kostic and Aigner (2007); Lunt and Bridge (2004); Nobes et al. (2001); Rice et al. (2009); Słowik (2011)

Delta Barnhardt and Sherrod (2006); Gibbard et al (2012); Gutsell et al. (2004)

Coastal Bennett et al. (2009); Bristow and Pucillo (2006); Buynevich et al. (2010); Clemmensen et al. (2012); Nielsen and Clemmensen (2009); Olson et al. (In Press); Pascucci et al. (2009); Tamura et al (2011)

Aeolian Bristow et al. (2007, 2010); Buynevich et al. (2010); Clemmensen et al. (2012); Tamura et al. (2011); Vriend et al. (2012)

Peatland Comas et al. (2005, 2011); De Oliveira et al. (2012); Kettridge et al. (2008); Lowry et al. (2009); Plado et al. (2011); Rosa et al. (2009)

Faults Bhosle et al. (2007); Christie et al. (2009); Malik et al. (2010)

3 Martin Robinson et al.


Figure 1. (a) The propagation of an EM wave in subsurface material (adapted from Neal, 2004); (b) Fixed offset profiling (adapted from Annan, 2001); (c) Common mid-point (CMP) sounding (adapted from Annan, 2001). GPR can quickly be deemed an unsuitable technique if the target is beyond the depth range of the GPR system. For GPR to work effectively, the target must exhibit electric properties (dielectric constant and electrical conductivity) which contrast with the host subsurface. The strength of an EM reflection is proportional to the magnitude of this contrast, with the amount of energy reflected

being given by the reflection coefficient (𝑅), expressed as (Equation 3):

𝑅 = !! ! !!!! ! !!

(Equation 3)

In which 𝑣! and 𝑣! are the velocities for layers 1 and 2 i.e the target and the host subsurface (Neal, 2004). In all cases, the value of 𝑅 will be between +1 and -1, with values further from zero representing greater differences in electrical properties. Table 2 can be used as a guide to determine the suitability of GPR for a particular study, with velocity values for typical subsurface mediums being provided. The environment under investigation needs to be examined before performing a survey. Metallic objects and major structural features within the vicinity of the survey location can seriously affect GPR results (Annan and Cosway, 1994).

Survey Design This discussion on survey design focuses on single fold, fixed-offset reflection profiling (Figure 1b), the most common mode of GPR data collection, however the design factors can be applied to other survey types (Jol and Bristow, 2003; Annan, 2005). The step-like procedure involves moving a signal transmitter and receiver, in a fixed antenna geometry, over the surface in repetitive steps. This mode is preferred in studies that require high spatial-horizontal resolution. Continuous data collection is not recommended for detailed sedimentary investigations as the movement during collection can smear the data, creating problems when locating a specific subsurface feature. Continuous data collection is more appropriate for reconnaissance surveys (Jol and Bristow, 2003). Several parameters need to be defined when designing a single fold, fixed-offset reflection survey.

Radar Frequency: Vertical Resolution Antenna frequency is a significant factor in survey design, determining the resolution and depth of penetration (Baker et al., 2007). Low frequency waves tend to penetrate deeper into the subsurface as they are not as easily



Table 2. Typical dielectric constant, electrical conductivity, velocity and attenuation values of common subsurface materials (Leckebusch, 2003; rows labelled * from Annan, 2005)

Material Dielectric Constant

Electrical Conductivity (mSm-1)

Velocity (m ns-1)

Attenuation (dB m-1)

Air 1 0 0.3 0

Salt water 80 3000 0.033 600

Fresh water 80 0.5 0.033 0.1

Ice* 3-4 0.01 0.16 0.01

Granite, dry 5 0.01 0.13 0.01

Limestone* 4-8 0.5-2 0.12 0.4-1

Shales* 5-15 1-100 0.09 1-100

Sand, dry 5 0.01 0.13 0.01

Sand, wet* 20-30 0.1-1.0 0.06 0.03-0.3

Clay, wet 10 500 0.095 300

Soils:

sandy, dry 2.6 1.4 0.19 1

sandy, wet 25 69 0.06 23

clayey, dry 2.5 2.7 0.19 3

clayey, wet 19 500 0.07 200

frozen 6 0.1 0.12 0.1

attenuated as high frequency waves. The ability of a GPR system to resolve fine subsurface features however deteriorates at lower frequencies, meaning the selection of the optimal antenna frequency is a trade-off between the desired resolution and the achievable penetration depth (Harari, 1996). As a rule of thumb, it is better to trade resolution for depth, with high resolution being useless if the target cannot be detected (Annan, 2005). An antenna does not transmit EM waves at a single frequency but across a frequency spectrum, with the higher frequencies in the range being preferentially attenuated as they propagate through the subsurface. This results in the return centre frequency, the most common frequency detected by the receiving antenna, typically being lower than the nominal centre frequency of the transmitting antenna. More realistic vertical resolution estimates are consequently obtained using the return centre frequency (Neal, 2004). The preferential attenuation of high frequency radar waves also causes a decrease in resolution down the GPR profile even when no changes in the velocity exist.

The majority of sedimentary studies use antenna with frequencies between 50 and 500MHz (Jol and Bristow, 2003). Table 3, based on the assumption that the required spatial resolution is approximately 25% of the target depth, can be used as a simple guide to determine a suitable frequency. It is important to understand that the depth of penetration is highly dependent on the clay content of the medium, with clay significantly attenuating EM waves and limiting the investigation depth (Bristow, 2013). The information presented in Table 2 therefore should not be a substitute for thorough survey planning, with the depth of penetration being subject to the soil type (Annan, 2001). Table 3. Guideline frequency values (Annan, 2001)

Depth (m) Centre Frequency (MHz)

0.5 1000 1.0 500 2.0 200 5.0 100

10.0 50 30.0 25 50.0 10



Stratigraphy in sedimentary environments occurs at scales ranging from 10-3 to 101 m both horizontally and vertically (Jol and Bristow, 2003). Identifying a suitable resolution is often a vital component of many subsurface environmental studies. Vertical resolution is usually considered to be approximately one quarter of the radar signal wavelength in the subsurface, with wavelength being a function of antenna frequency and the velocity of the signal (Bristow, 2009). This is shown by Equation 4.

λ = !! (Equation 4)

λ = Wavelength V = Velocity f = Frequency

Bristow (2009) provides theoretical values for GPR resolution in sands based on Eq.4 (Table 4). For example, a 100MHz transmitter used to investigate sand saturated with fresh water (velocity of 0.06m ns-1) will produce a wavelength of 0.6m and a resolution of 0.15m (one-quarter of the wavelength value). The approximate velocity values presented in Table 2 can be used with a desired resolution value to identify a suitable GPR frequency for a particular study. Various antennas frequencies can be employed to effectively image the subsurface using GPR (Jol and Bristow, 2003). The use of multiple frequencies depends entirely on the research question, with the time and cost of swapping several different antennas, as well as processing data collected at different frequencies, needing to be considered. Time Window The required time window (W) is estimated using the following equation (Equation 5):

W = 1.3 2×DepthVelocity (Equation 5)

with the maximum depth and minimum velocity likely to be observed being used. The equation accounts for uncertainty in velocity and depth variations by increasing the estimated time by 30% (Annan, 2005). Table 2 can be used as a guide if no information on the electrical properties of the survey site is available.

Temporal Sampling Interval The time interval between points on a recorded waveform is another parameter which needs to be considered when designing a survey. Annan (2005) indicates that as part of a good survey design, the sampling rate should be approximately six times the centre frequency of the antenna being used, with the following expression being utilised to calculate suitable sampling intervals (Equation 6):

𝑡 = !"""!!

(Equation 6) In which 𝑡 is the maximum sampling interval (ns) and 𝑓 is the centre frequency (MHz). Table 5 indicates the calculated sampling interval for a variety of antenna frequencies. The sampling values presented in the table should only be exceeded when data volume and acquisition speed are more important than the integrity of the data (Annan, 2001). The consequence of using small sample values is that random signals (noise) are given more influence on the trace. An increase in antenna frequency requires an increase in the sample rate due to the preferential attenuation at high frequencies. Increases in sample rate should be based on the Nyquist principle - the greatest vertical resolution that can be expected is one-quarter the wavelength (Jol and Bristow, 2003).

Table 4. Theoretical values for GPR resolution in different sands (Bristow, 2009). Note: Velocity value for dry sand differs slightly from figure provided in Table 1

Antenna Central Frequency (MHz)

Theoretical Resolution (m) for Saturated Sand

(0.06 m ns-1)

Theoretical Resolution (m) for Damp Sand

(0.1m ns-1)

Theoretical Resolution (m) for Dry

Sand (0.15m ns-1) 50 0.3 0.5 0.75

100 0.15 0.25 0.375

200 0.075 0.125 0.1875

400 0.0375 0.0625 0.09375



Table 5. Suitable sampling intervals and corresponding antenna frequencies (Annan, 2001)

Antenna Centre Frequency (MHz)

Maximum Sampling Interval (ns)

10 16.7

20 8.3

50 3.3

100 1.67

200 0.83

500 0.33

1000 0.17

Each trace in a GPR profile should be vertically stacked to enhance the return signals, with the running average of a number of radar transmissions being taken (Jol and Bristow, 2003). If a GPR system was programmed to collect 512 samples for each trace, with the system also being set up to stack 16 sequential traces into one record, then at least 8192 pulses (512 pulses multiplied by 16) would need to be transmitted for every recorded reflection trace (Conyers, 2013). This stacking process minimises random signals (noise) and emphasises persistent signals (reflections). An increase in trace stacking however will increase survey time (Jol and Bristow, 2003).

Step Size: Horizontal Resolution Step size, the distance between each data collection point made along a GPR profile, is an important part of the survey design process (Jol and Bristow, 2003). A minimum step size for each antenna frequency is usually suggested by the manufacturer (Bristow, 2009). The extent of the target however outweighs the guidelines proposed by the manufacturer, with the object of interest needing to be resolved both horizontally and vertically. These values are based on the Nyquist sampling interval, one-quarter of the wavelength in the subsurface and expressed as (Equation 7):

n! =!"!√! (Equation 7)

n! = Sampling Interval (m) f = Antenna Centre Frequency (MHz) K = Dielectric Constant

A maximum step size of one metre, often less (0.1-0.5m) depending on antenna frequency, should be used in sedimentary studies to provide detailed horizontal resolution of structures (Jol and Bristow, 2003). A typical survey performed with 100MHz antennas should have a step size of 0.25m, however a larger step size can be used when the subsurface comprises continuous horizontal layers. GPR data will not adequately define steeply dipping reflectors if the step size is too large (Annan, 2005).

Antenna Separation The majority of GPR systems have separate antennas for transmitting and receiving (bistatic operation), with some antennas having a fixed separation while others can be varied (Annan, 2005; Jol and Bristow, 2003). Antenna separation should be as small as possible based on the needs of the survey and the wavelength of the antennas. The depth resolution of targets decreases as the distance between antennas increases, however this effect is not significant until the antenna separation nears half of the target depth (Jol and Bristow, 2003). A safe antenna separation, if there is very little site information available, is 20% of the target depth (Annan, 2005).

Line Spacing and Orientation Line spacing is a compromise between the amount of detail required for the survey and the extent of the survey area. To avoid spatial aliasing, where the target is not correctly resolved, the line spacing is required to be approximately one quarter of the extent of the structure being investigated (Bristow, 2009). Survey lines should be arranged, whenever possible, in a grid, with lines running parallel and perpendicular to the expected dip direction of the structure (Jol and Bristow, 2003). Bristow (2009) indicates that images of sedimentary structures are improved if data is collected in the up-dip direction. Closely spaced lines give sufficiently dense coverage to produce 3D block models of the subsurface (Jol and Bristow, 2003). Grasmueck et al. (2005) recommend a quarter-wavelength spatial sampling (step size) as the minimum requirement for a full resolution 3D survey.



Figure 2. GPR profile using unshielded 200MHz antennas with 0.1m step size, dewow and SEC gain in a parkland setting (a) with trees and a fence, includes reflections from above (b) and beneath the ground (c) with interpretation (d) see text for details



Antenna Shielding When a GPR antenna directs EM energy into the subsurface, a portion of the energy is lost to the air. These EM waves, as they propagate through the air, can encounter an object or planar surface with a different dielectric constant, resulting in part of the signal being reflected to the receiver (Neal, 2004). Thus, not all reflections on radar profiles necessarily originate from features within the subsurface which can be seen in (Figure 2). The GPR profile in Figure 2 includes reflections from objects above and beneath the ground. These can be readily distinguished by velocity analysis fitting hyperbolic curves to observed hyperbolic reflections. The hyperbolas from the trees with metal tree guards indicate a velocity around 0.3mns-1 (Figure 2b) which is consistent with the velocity of the GPR signal through the air at the speed of light (0.299 mns-1). The tighter hyperbola shown in Figure 2c indicates a velocity of 0.064 mns-1, a typical velocity for water saturated sediments. The gently inclined planar reflection picked out in blue in Figure 2d could be mistakenly interpreted as dipping bedrock strata but actually comes from a fence between 12.5 and 15.5 m from the GPR profile as shown in Figure 2a. Shielding can be used to reduce these unwanted surface reflections. However, shielded antenna never fully eliminate airwaves, therefore shielded antenna data should be treated with caution. Operators should always look for potential sources of airwave in the area of study. Profile lines - run towards and away from a possible source to evaluate if the object is scattering energy (Annan, 2005).

Velocity Sounding Design The velocity of an EM wave needs to be determined to convert two-way travel-time into depth. This value can be obtained by the basic process of probing or excavating to determine depth to a known reflector, then using the depth and measured reflection time to calculate the velocity (Jol and Bristow, 2003). An example of this “ground truth” approach is shown in Figure 3 where a CMP survey on the top of the outcrop indicated a surface velocity of 0.13mns-1 (Jol et al., 2003). However, correlation with a measured section of the outcrop indicates an average velocity at depths up to 30m is 0.1mns-1.

Common mid-point (CMP) sounding (Figure 1c) and wide angle reflection and refraction (WARR) surveys estimate signal velocity by increasing the separation between the transmitter and receiver in steps at a fixed location and measuring the change of the two-way travel-time to reflections (Jol and Bristow, 2003; Annan, 2005). CMP surveys are generally considered to be more precise than WARR surveys and should be the first survey completed on arrival at a site (Jol and Bristow, 2003). Jol and Bristow (2003) suggest to start a CMP survey with a minimum antenna separation then move each antenna out by 5 or 10cm until no signal is returned. This however is only appropriate when performed above a horizontal reflector in multiple directions. Some software programs can measure the angle of hyperbolic reflections as a proxy for the moveout encountered in CMP, calculating velocity by fitting curves to observed hyperbolas as illustrated in Figure 2b and Figure 2c.

Data Processing GPR data processing can be a daunting task for new users. This section provides good practice guidelines to ensure consistent, efficient and realistic processing. A brief overview of several basic processing techniques is provided. It is important to keep processing simple. The majority of GPR data collected in sedimentological environments only requires minimal processing to allow interpretation (Jol and Bristow, 2003). The temptation to overprocess should be avoided as more sophisticated processing methods are likely to introduce bias and potential artefacts into the data (Cassidy, 2009). The user should take a systematic and consistent approach to the processing sequence, accurately recording all processing steps and parameters. GPR utilises many aspects of seismic data analysis that have been developed to industry-standard. Although not all seismic processing methods can be applied to GPR data, the majority can be used directly (Annan, 2005). Basic processing steps, usually directly applied to the raw data, typically take the form of trace editing, filtering or data



correction, introducing minimal user bias. Most, if not all, of these steps need to be used to make a basic interpretation. It must

be remembered that no amount of processing will save poor quality data (Cassidy, 2009).

Figure 3. 50 MHz GPR profile of the Navajo Sandstone, a Jurassic cross-stratified aeolian sandstone at Zion National Park in Utah (adapted from Jol et al., 2003). Reflections of cross-stratification and the bounding surfaces between them are clearly imaged by the GPR to a depth of almost 30m. A photograph of the outcrop is shown in Figure 8. Beneath 30m depth the arc shaped crossing reflections are reflections through the air from the canyon walls.

Data Editing Data editing, often the most time consuming part of a processing sequence, involving tasks such as data reorganisation, data file merging and updating background information (Annan, 2005). The occurrence of inevitable errors during data acquisition means some traces may need to be reversed, merged or omitted. The maintenance of data is important when trying to obtain good quality interpretations (Cassidy, 2009). Time-zero Correction Traces need to be adjusted to a common time-zero position before processing can be

applied. This usually involves setting a particular rule such as the time-zero position is the airwave first break point or the first negative peak of the trace. Successful realignment causes all reflections beneath to become correctly aligned. Time-zero correction is therefore often the first data processing step performed (Neal, 2004). The processing software often performs this automatically (Cassidy, 2009).

Dewow Filtering Dewowing is a basic processing step which uses temporal filtering to remove very low frequency components from the data (Annan, 2005). It reduces the data to a mean zero level, with the majority of modern GPR



systems applying dewow to each trace automatically (Cassidy, 2009). For advice on manual dewow correction see Cassidy (2009).

Topographic Correction Topographic correction is required to place the GPR data within its correct spatial context (Cassidy, 2009). When the surface and subsurface stratigraphy are horizontal, elevation static corrections can be used to topographically correct data, repositioning the time zero in the vertical axis and adjusting reflections accordingly (Bristow, 2009). Dipping reflections can be restored to the correct dip by migrating the data, then applying the static correction for topography (Bristow, 2009). For information on advanced methods, such as migration, see Cassidy (2009).

Gain The next basic processing step is to select a gain function for the data. Gains improve the visual form of the GPR sections, with most techniques altering the data structure in some form. It is therefore important to understand the effects of gain functions. Radar signals are rapidly attenuated as they propagate through the subsurface, making events from greater depths more difficult to discriminate (Annan, 2005). Gains enhance the appearance of later arrivals due to the effect of signal attenuation and geometrical spreading losses (Cassidy, 2009). There are several different types of gains including constant gain, exponential gain, exponential gain compensation (SEC) and automatic gain control (AGC), with each function having different characteristics. Gain functions can be easily changed, usually by altering the time window (a region of the trace in time), the gain function (linear, exponential, user-defined, etc) and the maximum gain allowed (Cassidy, 2009). Constant, linear or exponential gains are systemically applied gain functions that have a specific mathematical operator that is defined by the user or system automatically (Cassidy, 2009). Automatic gain control (AGC) and spherical and exponential gain compensation (SEC) are two of the most popular types of gain

function. AGC equalises the amplitudes down each trace, making it ideal for monitoring stratigraphic horizon continuity as well as the continuity of other reflections. This function however eliminates all amplitude information (Jol and Bristow, 2003). SEC attempts to emulate the variation in signal amplitude as it travels through the subsurface (Annan, 2005). Unlike AGC, SEC retains the relative amplitude information, with the reflections representing the true strength of the returned signal (Cassidy, 2009). This however only really applies if you have equivalent gains at the same depth.

Filtering Filters are applied to remove system or human-induced noise and improve the visual quality of the GPR data (Cassidy, 2009). There are many different types of filters, from simple band-pass filters to sophisticated domain and transform filters. Simple filters are usually very effective at removing high/low-frequency noise, while sophisticated functions are more appropriate for specific problems. Filters can be applied before or after gains however pre-gain filters operate on the data in its truest form (Cassidy, 2009). Filters can average down the trace (temporal) or from trace to trace (spatial), smoothing the data and removing high frequency noise (Bristow, 2013). Temporal filters are used to remove noise at frequencies that are higher or lower than the main GPR signal, ultimately acting as clean-up filters which make the GPR section visually better (Jol and Bristow, 2003). Spatial filtering is applied to either suppress or emphasise specific features. These filters are often used to remove the strong air/wave response and ringing from GPR data (Cassidy, 2009). Low pass spatial filters are suitable when trying to identify sedimentological features of interest, such as bedding, which are usually spatially extensive and low-angled (Cassidy, 2009). A high pass spatial filter has the opposite effect in that it suppresses flat-lying reflections and emphasises dipping events (Annan, 2005). High pass filters (frequency domain filters) allow higher frequency components to pass while removing low frequency components - low pass filters do the opposite. High-pass and low-pass filters can be combined in



band-pass filters, letting through frequencies either side of the peak frequency of the transmitted signal (Bristow, 2013). More information and advice on when to use certain filters can be found in Cassidy (2009).

Background Subtraction Background removal is one of the most common processing steps applied to GPR data (Annan, 2005). It often takes the form of a high pass filter or an average trace removal (a form of spatial filtering). This step allows subtle weaker signals to become visible in the processed section (Annan, 2005). Background subtraction is usually not necessary, therefore if data always requires background removal the equipment used to collect it may be flawed (Annan, 2005). For information on advanced processing methods see Cassidy (2009).

Visualising Processed Data The processed GPR data only becomes useful when it is available in image format (Daniels et al., 1988). It is traditionally presented as a radargram, a pseudo cross sectional image comprising a horizontal distance axis against a vertical TWT axis (Figure 4; Noon and Narayanan, 2002). These can be difficult to interpret, with several closely spaced point scatterers producing reflection hyperbolas which blur the radargram (Brunzell, 1999). The identification of multiple targets within a radargram unavoidably introduces subjectivity to the data analysis process (Daniels, 2004).

A selection of GPR profiles, collected from a range of depositional environments, has been provided. Figure 5 shows an interpreted 200 MHz GPR profile across a normal fault at Piano di Pezza in the Italian Apennines. The processed GPR profile shows a strong sub-horizontal reflection that dips gently towards the north. At its southern end, this reflection is at a depth of approximately 0.5m below the surface. Towards the northern end of the profile, the deep purple reflection is at a depth of approximately 6m from the surface at its deepest part, approximately 7m south of the footwall scarp. The change in depth to the reflector is indicative of subsidence adjacent to the fault and the accumulation of sediment in the hanging wall of the fault. Breaks in the reflections where there is a vertical offset are interpreted as faults; the displacement of the reflection is of the order of 1m indicating a fault displacement of a similar magnitude. Between the northward dipping reflections there is a “wedge” of weakly defined reflections which thins towards the south, away from the main fault scarp. Offsets in the reflections indicate approximately 0.5m vertical displacement across faults. The faulting identified on the GPR profile is dominated by vertical offset across synthetic and antithetic normal faults within and parallel to the main fault zone and this is consistent with the published trench profile (Pantosi et al., 1996). The main zone of deformation is localised within 8m south of the fault on the trench and GPR profiles. For additional information on this study see Jewell and Bristow (2004).

Figure 4. Radargram illustrating decrease in depth of peat



Figure 6 shows 200 MHz GPR profiles across sand dunes a linear dune in Namibia to investigate the internal structure of linear dunes from Bristow et al. (2000). Figure 7 displays a 200 MHz GPR profile across a gravel beach ridge, Waitaki coast, South Island, New Zealand. The profile shows inclined reflections interpreted as sedimentary layers within the beach ridge. The reflections are divided into two radar facies shaded orange and pink. Seaward dipping reflections are interpreted to be formed by beach (foreshore) progradation (Orange), while landward dipping reflections are interpreted as backshore washover deposits (Pink). Radar sequence boundaries are picked at horizons where there are reflection terminations or truncations marking erosion of the foreshore during storms. Some of these surfaces can be correlated with surfaces within the washover deposits which are interpreted to have been formed during coeval storm events. For further information see Dickson et al. (2009). As processing power has advanced, GPR data has been increasingly presented in the form of 3D models and time/depth slices (Annan, 2009). 3D surveying involves collecting data along closely spaces lines, with sufficiently dense data coverage allowing the generation of 3D block models (Jol and Bristow, 2003). These 3D models allow the user to interpret the depositional structure of stratigraphic units in greater detail than was possible using sparsely spaced 2D profiles (Jol and Bristow, 2003). Although 3D surveying can provide the best visualisation of the subsurface, the collection of 3D data is time consuming, with a considerable amount of time being required to accurately record the position and elevation of the data points (Serma and Setan, 2009; Bristow, 2009). This is why, in general, only small areas are surveyed. In the case of Figure 8, a grid 10m x 7m was surveyed with 200MHz antennas which imaged to a depth of around 12m. This can be compared with the depth of penetration of almost 30m achieved on the same outcrop with 50MHz antennas (Figure 9). With a longer wavelength the 50MHz antennas penetrate much deeper but with an

associated loss of resolution. See figures on outcrop for scale and log on Figure 3 for depth and bedding thickness. Modified from Jol et al. (2003).

Interpretation The interpretation of a GPR profile is inherently subjective, dependent on the knowledge, skill and experience of the user (Annan, 2005). GPR profiles should receive preliminary interpretation once collected to determine if resurveying is required (Jol and Bristow, 2003). After the data has been processed, an interpretation, independent of sedimentary models and other datasets such as cores, should be completed. Many interpretations based on models and other data can overlook key components that are visible in the GPR profile (Jol and Bristow, 2003). At the surface the earliest arrivals are the direct transmissions from the transmitter to the receiver through the air at the speed of light, the air wave, after this comes the direct signal through the ground or along the surface, the ground wave. These two direct signals form a pair of black and white lines along the top of all GPR profiles. They cover any reflections approximately equivalent to a depth of half the antenna spacing. Reflections from the subsurface appear beneath the air and ground waves. The GPR signal is not a simple sine wave but has a more complex waveform and as a result a single reflector can appear as a pair of reflections. This is sometimes overlooked resulting in over interpretation of GPR profiles. Data should initially be printed out, with major reflections being identified and traced around. The next step is to identify and characterise (as downlap, onlap, offlap, etc) reflection terminations, constructing a chronology of sedimentary deposits (Jol and Bristow, 2003). Figures 7 and 9 show examples with bounding surfaces picked out in red and reflection terminations marked by small black arrows. The ability to identify significant radar reflection patterns allows the location of features such as bedding planes, bedsets and cross-stratification to be determined (Jol and Bristow, 2003).



Figure 5. An interpreted 200 MHz GPR profile across a normal fault at Piano di Pezza in the Italian Apennines (modified from Jewel and Bristow, 2004).



Figure 6. 200 MHz GPR profiles across sand dunes a linear dune in Namibia to investigate the internal structure of linear dunes (Bristow et al., 2005).



Figure 7. 200 MHz GPR profile across a gravel beach ridge, Waitaki coast, South Island, New Zealand (Dickson et al. 2009). Although saline sea-water will normally attenuate the GPR signal, this coarse grained, well drained gravel beach combined with high rainfall means that sea-water is flushed from the beach sediments permitting imaging of beach progrades and washovers.



Figure 8. Site photograph accompanied by a 10m x 7m grid surveyed with 200MHz antennas, figures for scale (adapted from Jol et al., 2003). The depth of penetration (approx. 12m) and resolution of <0.2m can be compared with the 50MHz profile of the same outcrop shown in Figure 3 where depth of penetration is approx. 30m with resolution approx. 1m.



Figure 9. Image indicates how a GPR profile (a) can be interpreted to identify radar sequence boundaries (b) to derive a relative chronology (c) and select sites for optical dating to constrain rates of dune migration (Bristow et al., 2005).

(a)

(c)

(b)



Figure 9 shows a GPR profile across a sand dune in the Namib Sand Sea from Bristow et al. (2005). The top panel (Figure 9a) shows the GPR profile with topographic correction. The second panel (Figure 9b) shows reflection terminations marked by small arrows with truncation where arrows point up, and downlap where arrows point down. The radar sequence boundaries which are picked out in red mark breaks in deposition. On the

basis of superposition and cross-cutting relationships a relative chronology is derived numbers 1 to 22 in the yellow shaded section (Figure 9c). The interpreted GPR profile has been used to select sample points for dating in order to determine the rate of migration of the sand dune (Bristow et al., 2005). Another common interpretation strategy is to identify reflection patterns with similar

Figure 10. Annotated 100MHz GPR profile collected across part of a sand bar in the Jamuna / Brahmaputra River, Bangladesh (adapted from Best et al., 2003).



geometry which are termed radar facies (Jol and Bristow 2003). An example of radar facies interpretation is shown in Figure 10. Figure 10 shows a 100MHz GPR profile collected across part of a sand bar in the Jamuna/Brahmaputra River, Bangladesh. Three different reflection patterns have been identified and interpreted as radar facies, including: discontinuous concave reflections from sets of trough cross-stratification produced by dune bedforms migrating across the top of the sand bar; inclined tangential reflections from the lee-side slipface on the downstream margin of the bar migrating obliquely across the channel; continuous undulating reflections from scour and fill at the base of the channel (Best et al., 2003). The source of all signals should always be questioned. Novice users tend to incorrectly assume that all GPR responses originate from the subsurface. Airwave events, features associated with energy that propagates through the air and gets reflected back to the GPR receiver, can originate from above ground objects including fences, telephone poles and overhead wires (Annan, 2005). Such lateral reflections of the airwave can appear as hyperbolic features on a GPR profile (See Figure 2). Survey areas should be checked for potential sources of airwave events. If a potential source is found, profile lines should be run towards and away from the object to assess its effect on the GPR output. Users should always treat shielded antenna data with doubt as antenna shielding is never fully effective (Annan, 2005). Case Study: Determining Peat Thickness using GPR Aim of the Study GPR was used to determine the thickness of peat at a site in Northern Ireland. The aim of the study was to rapidly acquire continuous peat thickness data for a profile located on an upland blanket bog (a peatland up to 6m deep) situated approximately 9km north-west of the town of Dungiven (Co.Derry). The site, a gently sloping hill facing the north and north-west, has an average topographical height of approximately 210mAOD. The underlying geology of the bog is psammite and pelite, with a superficial geology of sandy silt and sandy clay.

Existing studies (Table 1) suggest that GPR can be used effectively to identify the subsurface contact point between the base of the peat and drift material.

Survey Design 50 MHz antennas were used to ensure the maximum depth of the peat would be identifiable (~6m). Although the average peat thickness for the site was approximately 1.5m, blanket bogs can reach depths of 6m. The employment of an antenna frequency of 100MHz, reaching an approximate depth of 5m (Table 3), may have resulted in the target (peat and subsoil interface) not being detected. The survey was completed using a MALÅ 50 MHz Rough Terrain Antenna (RTA). The flexible snake-like design of the antenna (comprising both a transmitter and receiver) provides optimum results in difficult environments, with the equipment allowing the operator to perform a survey without having to clear an access route. This system is therefore ideal for surveying peatlands affected by turf cutting (removal of peat for fuel or horticulture). Continuous data collection was used in this survey, with a step size therefore not being specified. A time window of 405ns was employed, with the time trigger interval being every half second and traces being stacked by 16 - settings were selected based on the experience of GPR users at Queen’s University Belfast. The antennas of the MALÅ system have a fixed separation meaning the EM wave velocity could not be determined using CMP sounding. Velocity was determined through the manual probing of the transect line, a process providing accurate information on the depth of the subsurface reflector. A series of connected probes were driven into the peat until a significant change in resistance was encountered. The peat probe was then removed and measured using a tape. Using the thickness of the peat and the basal reflection on the GPR profile, a velocity of 0.037mns-1 was calculated. Due to the survey environment, it was not possible to collect data in a straight line. Differential global positioning system (dGPS)



Figure 11. (a) Unprocessed radargram of case study transect; (b) Processed radargram of case study transect; (c) Processed radargram of

case study transect with base of peat labelled.

(a)

(c)

(b)



points were collected approximately every 20m along the profile using a Leica SR530 rover system, with the degree of sinuosity in the survey line being investigated in ArcGIS. It was determined that each dGPS point, on average, deviated from the predetermined survey line by approximately 2.50m, with the greatest deviation being 8.07m. This degree of sinuosity in the profile was deemed acceptable, considering the extent of large-scale vegetation and safety issues related to historical peat cutting.

Data Processing Minimal processing was applied to the collected data before interpretation. De-wowing and time-zero correction were automatically applied to the data. Automatic gain control (AGC) was employed,

accentuating stratigraphic horizon continuity in the radargram. Background subtraction, a common processing step, was then applied, removing noise from the data. Down-the-trace averaging (running average) was used to reduce random noise, acting as a low pass filter.

Visualising Processed Data The data from the transect has been presented in the form of a radargram, with Figures 11a and 11b showing the unprocessed and processed data, respectively. Figure 11c shows the processed radargram, annotated with a broken line indicating the estimated base of the peat. Figures 12a and 12b show a subsection of the line, allowing reflections to be viewed in

Figure 12. (a) Processed radargram of case study transect section (200m - 400m); (b) Processed radargram of case study transect section (200m - 400m) with base of peat labelled

(a)

(b)



greater detail. Depth scale was determined using manual probe measurements. The radargram was printed, with the first continuous major reflection, the contact between the base of the peat and drift material, being traced. Although the data were not collected within close proximity of any metallic fencing or buildings, the profile was situated adjacent to a nearby windfarm, with the closest turbine being approximately 200m. The wind turbine, located approximately 150m along the profile, does not appear to have a noticeable impact on the radargram. This is due to the exponential decay of the emitted EM wave, with Table 3 clearly indicating that the footprint of a 50MHz antenna would not incorporate a wind turbine located over 200m from the survey line.

Summary The brief case study provided shows how GPR can be successfully utilised in environmental studies, providing a continuous profile of varying peat thickness values for a selected case study site. Minimal processing was successfully used to indicate the boundary between overlying peat and the superficial geology, with the visualised profile being very useful in peatland studies.

Advantages and Limitations A major advantage of GPR is that the scale of resolvable features can range from several centimetres to hundreds of metres using a selection of available antennas (Mellet, 1995). GPR systems have the ability to rapidly collect large amounts of continuous subsurface data (Hruska et al., 1999; Doolittle and Collins, 1995). Most GPR techniques also allow acquired data to be reviewed on-site for quality control purposes (Mellet, 1995). The ability to use remote non-contacting tranducers of radiated energy, as opposed to ground contacting types needed for seismic studies, is a significant advantage that GPR has over seismic techniques. A significant limitation of GPR is that the antennas transmit and receive EM energy in a complex 3D cone, with reflections originating from anywhere on the radar wave front (Neal, 2004). The impression that a

reflection on a 2D radar profile is obtained from directly beneath the survey point is incorrect. Out-of-plane reflections from isolated point reflectors and reflector surfaces (known as “side swipes”) are difficult to identify on reflection profiles (Neal, 2004). This can lead to errors in depth estimation and interpretation (Olhoeft, 1998) - see Figures 2 and 3 for examples of out of plane, airwave reflections. A major limitation of GPR is that the performance of a system can be seriously affected by the environment it is surveying, with conductive soils strongly attenuating EM waves. Rapp and Hill (2006) states that a 100MHz GPR can penetrate approximately 15m of dry sandy soil or sediment compared to as little as 1m of wet clayey soil. The rapid attenuation of EM waves by conductive fine-grained sediments can produce high frequency ringing which can subsequently obscure primary reflections (Neal, 2004). Moorman et al. (2003) indicates that problems can occur when using GPR to survey permafrost environments, with the noise created by diffractions from small ice lenses resulting in deeper structures being masked. Another problem is that limitations that are dependent on the applied data collection configuration cannot be overcome once the basic GPR survey data has been collected (Neal, 2004).

Conclusion GPR, a geophysical technique with multidisciplinary applications, has become a valuable tool in sedimentological studies. This article has provided good practice guidelines on the collection, processing and interpretation of GPR data. A variety of survey design factors were proposed including operating frequency, temporal sampling interval, step size and antenna separation. Basic data processing steps have been reviewed, with information also being provided on data visualisation and interpretation. Possible applications of GPR in geomorphological research have been presented, with a case study outlining how GPR has been used to measure peat thickness being provided.



Acknowledgements I would like to thank Dr. Alastair Ruffell and Dr. Jennifer McKinley for their guidance during the preparation of this article. Special thanks go to Prof. Charlie Bristow for providing some excellent GPR images from a range of environmental settings - they are greatly appreciated.

References Annan AP. 2009. Electromagnetic principles of ground penetrating radar. In Ground penetrating radar: theory and applications, Jol HM (ed). Elsevier: Amsterdam; 3-40.

Annan AP. 2005. Ground-penetrating radar. In Near surface geophysics, Butler DK (ed). Society of exploration geophysicists: Tulsa, Investigations in Geophysics 13; 357-438.

Annan AP. 2001. Ground penetrating radar workshop notes. Sensors and Software Inc: Mississauga.

Annan AP and Cosway SW. 1994. GPR frequency selection. In Proceedings of the fifth international conference on ground penetrating radar, Redman JD (ed). Waterloo Centre for Groundwater Research: Kitchener; 747-760.

Appleby JR, Brook MS, Vale SS and MacDonald-Creevey AM. 2010. Structural glaciology of a temperate maritime glacier: Lower Fox glacier, New Zealand. Geografiska annaler: series a, physical geography 92: 451-467.

Ashworth PJ, Sambrook Smith GH, Best JL, Bridge JS, Lane SN, Lunt IA, Reesink AJH, Simpson CJ and Thomas RE. 2011. Evolution and sedimentology of a channel fill in the sandy braided South Saskatchewan River and its comparison to the deposits of an adjacent compound bar. Sedimentology 58: 1860-1883.

Baker GS, Jordan TE and Pardy J. 2007. An introduction to ground penetrating radar (GPR). The geological society of America special papers 432: 1-18.

Barnhardt WA and Sherrod BL. 2006. Evolution of a Holocene delta driven by episodic sediment delivery and coseismic deformation, Puget Sound, Washington, USA. Sedimentology 53: 1211-1228.

Benediktsson ÍÖ, Ingólfsson Ó, Schomacker A and Kjær KH. 2009. Formation of submarginal and proglacial end moraines: implications of ice-flow mechanism during the 1963-64 surge of Búarjökull, Iceland. Boreas 38: 440-457.

Bennett MR, Cassidy NJ and Pile J. 2009. Internal structure of a barrier beach as revealed by ground penetrating radar (GPR): Chesil beach, UK. Geomorphology 104: 218-229.

Best JL, Ashworth PJ, Bristow CS and Roden J., 2003. Three dimensional sedimentary architecture of a large, mid-channel sand braid bar, Jamuna River, Bangladesh. Journal of sedimentary research 73: 516-530.

Bhosle B, Parkash B, Awasthi AK, Singh VN and Singh S. 2007. Remote sensing-GIS and GPR studies of two active faults, Western Gangetic Plains, India. Journal of applied geophysics 61: 155-164.

Bristow CS. 2013. Ground penetrating radar. In Treatise on geomorphology, vol 14, Shroder JF (ed). Academic Press: San Diego; 183-194.

Bristow CS. 2009. Ground penetrating radar in aeolian dune sands. In Ground penetrating radar: theory and applications, Jol HM (ed). Elsevier: Amsterdam; 273-298.

Bristow CS and Pucillo K. 2006. Quantifying rates of coastal progradation from sediment volume using GPR and OSL: the Holocene fill of Guichen Bay, south-east South Australia. Sedimentology 53: 769-788.

Bristow CS, Augustinus PC, Wallis IC, Jol HM and Rhodes EJ. 2010. Investigation of the age and migration of reversing dunes in Antarctica using GPR and OSL, with implications for GPR on Mars. Earth and planetary science letters 289: 30-42. doi:10.1016/j.epsl.2009.10.026

Bristow CS, Duller GAT and Lancaster N. 2007. Age and dynamics of linear dunes in the Namib Desert. Geology 35:555-558.

Bristow CS, Lancaster N and Duller GAT. 2005. Combining ground penetrating radar surveys and optical dating to determine dune migration in Namibia. Journal of the geological society 162(2): 315-322.

Bristow CS, Bailey SD and Lancaster N. 2000. The sedimentary structure of linear sand dunes. Nature 406: 56-59.



Brunzell H. 1999. Detection of shallowly buried objects using impulse radar. IEEE transactions on geoscience and remote sensing 37: 875-886.

Buynevich IV, Filho PWMS and Asp NE. 2010. Dune advance into a coastal forest, equatorial Brazil: A subsurface perspective. Aeolian research 2: 27-32.

Cassidy NJ. 2009. Ground penetrating radar data processing, modelling and analysis. In Ground penetrating radar: theory and applications, Jol HM (ed). Elsevier: Amsterdam; 141-176.

Christie M, Tsoflias GP, Stockli DF and Black R. 2009. Assessing fault displacement and off-fault deformation in an extensional tectonic setting using 3-D ground-penetrating radar imaging. Journal of applied geophysics 68: 9-16.

Clemmensen LB, Nielsen L, Bendixen M and Murray A. 2012. Morphology and sedimentary architecture of a beach-ridge system (Anholt, the Kattegat sea): a record of punctuated coastal progradation and sea-level change over the past ~1000 years. Boreas 41: 422-434.

Comas X, Slater L and Reeve A. 2011. Pool patterning in a northern peatland: geophysical evidence for the role of postglacial landforms. Journal of hydrology 399: 173-184.

Comas X, Slater L and Reeve A. 2005. Stratigraphic controls on pool formation in a domed bog inferred from ground penetrating radar (GPR). Journal of hydrology 315: 40-51.

Comas X, Slater L and Reeve A. 2004. Geophysical evidence for peat basin morphology and stratigraphic controls on vegetation observed in a northern peatland. Journal of hydrology 295: 173-184.

Conyers LB (2013) Ground-penetrating radar for archaeology (3rd edition). AltaMira Press: Plymouth.

Daniels DJ. 2004. Ground penetrating radar (2nd edition). The Institution of Electrical Engineers: London.

Daniels DJ, Gunton DJ and Scott HF. 1988. Introduction to subsurface radar. IEE proceedings f: radar and signal processing 135: 278–320.

De Oliveira MAT, Porsani JL, de Lima GL, Jeske-Pieruschka V and Behling H. 2012. Upper Pleistocene to Holocene peatland evolution in Southern Brazilian highlands as depicted by radar stratigraphy, sedimentology and palynology. Quaternary research 77: 397-407.

Degenhardt JJ. 2009. Development of tongue-shaped and multilobate rock glaciers in alpine environments – interpretations from ground penetrating radar surveys. Geomorphology 109: 94-107.

Dickson ME, Bristow CS, Hicks DM, Jol H, Stapleton J, Todd D. 2009. Beach volume on an eroding sand–gravel coast determined using ground penetrating radar. Journal of coastal research 25: 1149-1159.

Doolittle JA and Collins ME. 1995. Use of soil information to determine application of ground penetrating radar. Journal of applied geophysics 33: 101-108.

Gibbard PL, West RG, Boreham S and Rolfe CJ. 2012. Late Middle Pleistocene ice-marginal sedimentation in East Anglia, England. Boreas 41: 319-336.

Grasmueck M, Weger R and Horstmeyer H. 2005. Full resolution 3D GPR imaging. Geophysics 70: k12-k19.

Gutsell JE, Clague JJ, Best ME, Bobrowsky PT and Hutchinson I. 2004. Architecture and evolution of a fjord-head delta, western Vancouver Island, British Columbia. Journal of Quaternary science 19: 497-511.

Harari Z. 1996. Ground-penetrating radar (GPR) for imaging stratigraphic features and groundwater in sand dunes. Journal of applied geophysics 36: 43-52.

Hart JK, Rose KC and Martinez K. 2011. Temporal englacial water content variability associated with a rapidly retreating glacier. Earth surface processes and landforms 36: 1230-1239.

Holden J, Burt TP and Vilas M. 2002. Application of ground-penetrating radar to the identification of subsurface piping in blanket peat. Earth surface processes and landforms 27: 235-249.

Hruska J, Cermák J and Sustek S. 1999. Mapping tree root systems with ground-penetrating radar. Tree physiology 19: 125-130.



Irvine-Fynn TDL, Moorman BJ, Williams JLM and Walter FSA. 2006. Seasonal changes in ground-penetrating radar signature observed at a polythermal glacier, Bylot Island, Canada. Earth surface processes and landforms 31: 892-909.

Jewell C and Bristow CS. 2004. GPR studies in the Piano di Pezza area of the Ovindoli-Pezza Fault, Central Apennines, Italy: Extending palaeoseismic trench investigations with high resolution GPR. Proceedings of the Tenth International Conference on Ground Penetrating Radar, Delft, The Netherlands: 555-558 IEEEXplore.

Johnson BA and Carpenter PJ. In Press. Geophysical response of slackwater and sandy terrace deposits near Savanna, Northwestern Illinois. Environmental earth sciences. DOI 10.1007/s12665-012-1798-5

Jol HM and Bristow CS. 2003. GPR in sediments: advice on data collection, basic processing and interpretation, a good practice guide. In Ground penetrating radar in sediments, Bristow CS and Jol HM (eds). Geological Society: London, Special Publication 211; 9-28.

Jol HM and Smith DG. 1991. Ground penetrating radar of northern lacustrine deltas. Canadian journal of earth sciences 28: 1939-1947.

Jol HM, Bristow CS, Smith DG, Junck B and Putnam P. 2003. Stratigraphic imaging of the Navajo Sandstone using ground-penetrating radar. The leading edge, September 2003: 882-887.

Kettridge N, Comas X, Baird A, Slater L, Strack M, Thompson D, Jol H and Binley A. 2008. Ecohydrologically important subsurface structures in peatlands revealed by ground-penetrating radar and complex conductivity surveys [on-line]. Journal of geophysical research 113.

Kim KY, Lee J, Hong MH, Hong JK, Jin YK and Shon H. 2010. Seismic and radar investigations of Fourcade Glacier on King George Island, Antarctica. Polar research 29: 298-310.

Kostic B and Aigner T. 2007. Sedimentary architecture and 3D ground-penetrating radar analysis of gravelly meandering river deposits (Neckar Valley, SW Germany). Sedimentology 54: 789-808.

Langston G, Bentley LR, Hayashi M, McClymont A and Pidlisecky A. 2011. Internal structure and hydrological functions of an alpine proglacial moraine. Hydrological processes 25: 2967-2982.

Leckebusch J. 2003. Ground-penetrating radar: a modern three-dimensional prospection method. Archaeological Prospection 10: 213-240.

Leopold M, Williams MW, Caine N, Völkel J and Dethier D. 2011. Internal structure of the Green Lake 5 rock glacier, Colorado Front Range, USA. Permafrost and periglacial processes 22: 107-119.

Lowry CS, Fratta D and Anderson MP. 2009. Ground penetrating radar and spring formation in a groundwater dominated peat wetland. Journal of hydrology 373: 68-79.

Lunt IA and Bridge JS. 2004. Evolution and deposits of a gravelly braid bar, Sagavanirktok River, Alaska. Sedimentology 51: 415-432.

Malik JN, Sahoo AK, Shah AA, Shinde DP, Juyal N and Singhvi AK. 2010. Paleoseismic evidence from trench investigation along Hajipur fault, Himalayan Frontal Thrust, NW Himalaya: implications of the faulting pattern on landscape evolution and seismic hazard. Journal of structural geology 32: 350-361.

Mellet JS.1995. Ground penetrating radar applications in engineering, environmental management, and geology. Journal of applied geophysics 33: 157-166.

Monnier S, Camerlynck C, Rejiba F, Kinnard C, Feuillet T and Dhemaied A. 2011. Structure and genesis of the Thabor rock glacier (Northern French Alps) determined from morphological and ground-penetrating radar surveys. Geomorphology 134: 269-279.

Moorman BJ, Robinson SD and Burgess MM. 2003. Imaging periglacial conditions with ground-penetrating radar. Permafrost and periglacial processes 14: 319-329.

Murray T and Booth AD. 2010. Imaging glacial sediment inclusions in 3-D using ground-penetrating radar at Kongsvegen, Svalbard. Journal of Quaternary science 25: 754-761.

Neal A. 2004. Ground-penetrating radar and its use in sedimentology: principles, problems and progress. Earth science reviews 66: 261-330.



Nielsen L and Clemmensen LB. 2009. Sea-level markers identified in ground-penetrating radar data collected across a modern beach ridge system in a microtidal regime. Terra nova 21: 474-479.

Nobes DC, Ferguson RJ and Brierley GJ. 2001. Ground-penetrating radar and sedimentological analysis of Holocene floodplains: insight from the Tuross valley, New South Wales. Australian journal of earth sciences 48: 347-355.

Noon DA and Narayanan RM. 2002. Subsurface remote sensing. In Review of radio science 1999-2002 Stone WR (ed). Wiley-IEEE: USA; 535-552.

Olhoeft GR. 1998. Electrical, magnetic and geometric properties that determine ground penetrating radar performance. Proceedings of GPR’98, seventh international conference on ground penetrating radar. University of Kansas: Lawrence; 477-483.

Olson D, Kennedy DM, Dawe I and Calder M. In Press. Decadal-scale gravel beach evolution on a tectonically-uplifting coast: Wellington, New Zealand. Earth surface processes and landforms. DOI: 10.1002/esp.3233

Pantosti D, D’Addezio G and Cinti FR. 1996. Paleoseismicity of the Ovindoli-Pezza fault, central Apennines, Italy: a history including a large, previously unrecorded earthquake in the Middle Ages (860–1300 A.D.). Journal of geophysical research: solid earth 101(B3): 5937–5960.

Pascucci V, Martini IP and Endres AL. 2009. Facies and ground-penetrating radar characteristics of coarse-grained beach deposits of the uppermost Pleistocene glacial Lake Algonquin, Ontario, Canada. Sedimentology 56: 529-545.

Plado J, Sibul I, Mustasaar M and Jõeleht A. 2011. Ground-penetrating radar study of the Rahivere peat bog, eastern Estonia. Estonian journal of earth sciences 60: 31-42.

Powers MH. 1997. Modelling frequency-dependent GPR. The leading edge 16: 1657-1662.

Rapp G and Hill CL. 2006. Geoarchaeology: the earth-science approach to archaeological interpretation (2nd edition). Yale University Press: London.

Rice SP, Church M, Wooldridge CL and Hickin EJ. 2009. Morphology and evolution of bars in a wandering gravel-bed river; lower Fraser river, British Columbia, Canada. Sedimentology 56: 709-736.

Rosa E, Larocque M, Pellerin S, Gagné S and Fournier B. 2009. Determining the number of manual measurements required to improve peat thickness estimations by ground penetrating radar. Earth surface processes and landforms 34: 377-383.

Serma AIA and Setan H. 2009. Ground penetrating radar (GPR) for subsurface mapping: preliminary result. Geoinformation science journal 9: 45-62.

Słowik M. 2011. Reconstructing migration phases of meandering channels by means of ground-penetrating radar (GPR): the case of the Obra River, Poland. Journal of soils and sediments 11: 1262-1278.

Tamura T, Kodama Y, Bateman MD, Saitoh Y, Watanabe K, Matsumoto D and Yamaguchi N. 2011. Coastal barrier dune construction during sea-level highstands in MIS 3 and 5a on Tottori coast-line, Japan. Palaeogeography, palaeoclimatology, palaeoecology 308: 492-501.

Vriend NM, Hunt ML and Clayton RW. 2012. Sedimentary structure of large sand dunes: examples from Dumont and Eureka dunes, California. Geophysical journal international 190: 981-992.

Date post:	21-Aug-2019
Category:	Documents
Upload:	buicong
View:	213 times
Download:	0 times

· Title: Microsoft Word - 1.5.5_GPR.docx

Documents