The technical challenges and outcomes of ground-penetrating radar: A site-specific example from Joggins, Nova Scotia

The Carboniferous Joggins Formation is known for its complete succession of fossil-rich, coal-bearing strata, deposited in a fluvial meanderbelt depositional setting. Hence, the Joggins Formation outcrop is an excellent analogue for studying the 2D geological complexities associated with meanderbelt systems. In this research, a conventional ground-penetrating radar system was tested with the intent of imaging near-surface, dipping, strata of the Joggins Formation (potentially with subsequent repeats as annual erosion provides new visual calibrations). The survey was unsuccessful in its primary goal, and for future reference we document the reasons here. However, the overlying near-surface angular unconformity was successfully imaged enabling mapping of the approximately 8 m of overlying glacial till. A successful outcome would have allowed observations from the 2D outcrop to be extended into 3D space and perhaps lead to an increased understanding of the small (e.g., bedform baffles and barriers) and large (e.g., channel bodies) scale architectural elements, meanderbelt geometry, and aspect ratios. The study comprises a 42-line, 3.46 km ground-penetrating radar survey using a Sensors and Software pulseEKKO Pro SmartCart system. It was combined with a real-time kinematic differential global positioning system for the georeferencing of survey lines. The 50 MHz antenna frequency, with a 1 m separation, was chosen to maximize the depth of penetration, while still maintaining a reasonable resolution. The results show that many of the lines are contaminated with diffraction hyperbolae, possibly caused from buried objects near or under the survey lines or surface objects near the survey lines. A total of thirteen unique radar reflectors are described and interpreted from this work. The thick clay-rich soil overlying the Joggins Formation probably contributed to


Introduction
The use of 2D siliclastic outcrops for the study of reservoir analogues provides a wealth of knowledge relating to the interwell scale geometrical and petrophysical heterogeneities within a depositional system, which ultimately control permeability and porosity, and thus, the mobility and capacity of reservoir fluids [1]. Ideally, the 2D outcrop would be extended into the third dimension to allow for the development of a continuous model that would further help with interpretations. The issue that one is then faced with is how to best fill in the region behind the outcrop to create a 3D model [2]. One such method that has the potential of providing this data is ground-penetrating radar (GPR); a near-surface geophysical technique that can provide high-resolution images of ancient and modern sedimentary sequences, which can be used to improve the understanding of small (e.g., bedform baffles and barriers) to large (e.g., channel bodies) scale architectural elements, meanderbelt geometry, and aspect ratios, just to name a few [1][2][3][4][5][6][7][8][9][10][11].
GPR is a non-invasive and non-destructive remote sensing geophysical technique that is highly useful and versatile utilized in several different disciplines for the imaging and subsequent study of the shallow subsurface (e.g., [12,13]). It accomplishes this through the detection of electrical discontinuities by the generation, propagation, reflection, and reception of pulsed high-frequency electromagnetic energy (e.g., [12,13]). These discontinuities are directly related to water saturation, salinity, porosity, and mineralogical variations [3,14]. Ideal GPR results are typically achieved from clean, quartzose-rich clastic sediments that contain no clays or silts (e.g., [7,15]). Signal attenuation is a real concern when performing a GPR survey, with problems arising from concentrations of silt, clay, caliche, and moist saline conditions (e.g., [7,15,16]).
Here we provide the first comprehensive results from a primarily road-based GPR survey. In this study, a total of 42 GPR lines were collected over the Carboniferous-aged Joggins Formation of northern Nova Scotia, Canada, using one set of 50 MHz antennae arranged as a transmitter/receiver pair. The GPR system was combined with a Real-Time Kinematic (RTK) Differential Global Positioning System (DGPS) to provide a fully georeferenced group of survey lines with positional accuracy of approximately ±2 cm [17]. The goals of this survey were to image the dipping conformable strata of the fluvial-dominated Joggins Formation to identify sedimentary structures that could be correlated with a previously obtained lidar survey of the cliff face. This work was performed with the aim of providing constraining data on the reservoir architecture of the Carboniferous fluvial meanderbelt system of the Joggins Formation in 3D (outcrop + GPR). The results of the GPR study were to be used as inputs to stochastic models of the Joggins Formation with the purpose of understanding the inherent reservoir heterogeneity sensitivities of the analogous reservoir. The four main geometric measures include channel depth, channel width, sandstone thickness and channel-belt width. From those, four aspect ratios can be calculated, (1) channel depth versus sandstone thickness, (2) channel depth versus channel width, (3) channel-belt width versus channel depth, and (4) channel-belt width versus channel width. Gibling [18] documents the width and thickness of fluvial channel bodies from the geological record, including those measured from the 2D outcrop exposure of the Joggins Formation. Additionally, this study was carried out to test the applicability of the GPR system to provide high-resolution imaging of the dipping strata of the Joggins Formation, with the possibility that these images could be integrated with other outcrop (e.g., LiDAR) and subsurface data (e.g., drill core, well logs). The majority of the GPR data show strong diffraction hyperbolae, which is likely the result of above ground and subsurface objects. These objects could not be bypassed since most of the GPR lines were conducted on gravel/dirt roads traversing on top of the Joggins Formation strata. The objects that contaminate the radargrams must be understood and differentiated from the true sedimentary structures that were the purpose of this survey. There are also many other items that were unique to this survey that could potentially result in the radargrams being contaminated.
A search of previous research related to the Joggins Formation yields a vast number of publications that are either written directly about the Joggins Formation or mention the Joggins Formation in some capacity (e.g., [18][19][20][21][22][23]). According to a recent publication by Grey and Finkel [24], the bulk of the research occupies one of three major categories; a general geology category that includes sedimentology and stratigraphy publications (e.g., [21,22,25,26]); a paleobiology category that includes taxonomic discoveries and descriptions (e.g., [27][28][29][30][31]); and a paleoecology category (e.g., [20,32]). Despite the abundant research carried out in this area, there is a lack of research into the subsurface imaging of the Joggins Formation, particularly those that utilize ground penetrating radar. The exception dating back to the early 1960's, until approximately 2008, when numerous 2D seismic lines were collected in the onshore Cumberland Subbasin for the purpose of hydrocarbon resources exploration. In addition, several petroleum boreholes were drilled to test areas and structures of potential interest. Some of these wells penetrated the Joggins Formation strata. The Athol Syncline was the focus of a regional seismic study and appeared to show evidence of rapid subsidence within the Cumberland Subbasin resulting from Mississippian salt withdrawal at depth, allowing for thick sediment accumulations and preservation [26].

Study area
The community of Joggins is located approximately 230 km north of Halifax, Nova Scotia. Joggins and the corresponding outcrop lie alongside Chignecto Bay, a smaller bay within the grander Bay of Fundy. In this area, the tides ebb and flow some 13 m with each tidal cycle ( Figure 1). The mean annual temperature is approximately 6.0 °C, and the mean annual precipitation is 1154.8 mm. The ease of access, continuity and quality of the Joggins Formation exposure and the numerous road/grass surfaces over which a GPR survey could be completed are the main reasons behind its selection as a study site. The Joggins Fossil Cliffs (Joggins Formation) were nominated in 2008 as a United Nations Educational, Scientific and Cultural Organization (UNESCO) heritage site, together with six additional conformable formations (Ragged Reef, Springhill Mines, Little River, Boss Point, Claremont and Shepody) because of the exceptionally well-preserved rock outcrops and fossil assemblages that document life during the "Coal Age", a time when fertile forests and wetlands occupied the World's tropics [33].
Joggins and the nearby area have seen extensive coal mining that dates back to 1686 [34], continuing intermittently for over 200 years. During that time, elaborate underground mine workings were created, with many of the remnants (e.g., mine opening supports and railway line support timbers) visible in the cliff face from the intertidal zone (e.g., [34][35][36]). The surface development was also substantial, with timbers (rail track and support) and steel spikes still visible on the intertidal zone between Main Street and the Joggins Fossil Cliffs center. The remains of a wooden pier that existed for the load-out of coal onto ships during high tide for destinations throughout the Maritimes and New England [34]. The ground over which the GPR survey was conducted, has seen extensive, human-related activity, which needs to be accounted for when interpreting the data.

Figure 1.
Location map of Joggins, Nova Scotia. GPR data was collected from the areas outlined with red (basemap from [37]).

Overburden geology
Overburden (soil and glacial till) geology is important for GPR studies because of the direct impact it has on data collection, especially when attempting to image underlying bedrock. A review of borehole reports from the area describes the overburden as ranging in thickness from 1.5 to 20.1 m (average of 6.4 m). In the study area, the Joggins Formation is overlain by glacial till with a developed soil horizon on top for a total thickness of approximately 8 m (Figure 2). Joggins soils are grayish-brown, moderately fine-textured, and stony with poor internal drainage (e.g., [38,39]). The "A" horizon is at least 15 cm thick and characterized by pale gray to pale brown sandy loam and sandy clay loam with yellowish mottling, an indication of extended saturation and gleying (e.g., [38,39]). Underlying is a 10 to 25 cm thick yellowish and reddish interval of mottled material with a dullish brown matrix, manganese dioxide (MnO2) concretions and considerable free iron accumulation. The remaining "B" horizon begins at approximately 30 cm and continues to 60 cm depth. It is a compact, dense sandy clay loam with a weakening mottled texture and brown to dark grayish brown matrix [38]. Thin clay lines most of the voids, resulting in a low permeability zone. The underlying "C" horizon is also clay-rich, dense, and dull reddish-brown to grayish-brown with rare mottling. The detailed soil horizon characteristics are described by Nowland and MacDougall [38]. Glacial till is defined as a mass of unsorted debris deposited by a glacier and consisting of grain sizes ranging from boulders to clay [40]. The glacial till on which these soil horizons were formed is a grayish, silty clay loam that originated from the fine-grained grey and red sandstones, shales and mudstones of the Carboniferous coal measure beds [38]. A more recent study by Stea and Finck [41] names the till in this location as the Joggins Till. It is described as being a sandy silt that is dark yellowish-brown in colour with clasts composed of grey sandstones, shales, mudstones, and minor red sandstones and shales, limestone and coal [41].

Bedrock geology
The Maritimes Basin is comprised of ten onshore/offshore subbasins, of which the Cumberland Subbasin is one (Figure 3). It hosts numerous, well-known coal deposits, of which numerous seams and their associated mine workings are visible from the intertidal zone. The Joggins Formation is part of the Cumberland Group and with the Mabou Group forms a continuous 14.7 km long outcrop ( Figure 4) along the coast of Chignecto Bay [24]. At approximately 4500 km 2 , the Cumberland Subbasin is a fault-bounded depocenter containing some 7000 m of Late Devonian to Early Permian sediment (e.g., [42,43]). The subbasin occurs over areas of northwestern Nova Scotia and to a minor extent, regions of southern New Brunswick. It is positioned to the south by the Cobequid Mountains, to the west by the Caledonian Highlands and Westmorland Uplift, and to the east by the Antigonish Highlands (e.g., [43,44]). It is suggested by Browne and Plint [45] that the subbasin margins are comprised to the south by the North Fault, to the north by the Caledonia-Dorchester fault system, and to the west by the Harvey-Hopewell Fault. The northwestern basin margins, as suggested by Martel [46] are characterized by a laterally trending basal horst along the Hastings Fault.
A sequence of synclines occur in the basin, with the more significant examples being the Amherst, Athol, Scotsburn, Tatamagouche, and Wallace, in addition to a couple diapirism-related anticlines known as the Claremont-Malagash and Minudie, both being encircled by the aforementioned synclinal sequence [44]. According to Ryan and Boehner [44] the Cumberland Subbasin structural elements are correlated with basin growth features and include major synclines as well as growth and strike-slip faults. Those structural elements are either unrelated to or are indirectly related to salt tectonics and their related salt structures such as diapiric anticlines, diapirs, domes, and folds/faults related to salt movement [44]. The Cumberland Subbasin is considered a salt-withdrawal basin with both the slump features and movement of salt occurring concurrently with basin deposition.
The Joggins Formation has been interpreted to contain three stratigraphic facies; a well-drained floodplain facies that includes reddish siltstone, mudstone and sandstone with minor greyish mudstone, rare coal and limestone beds; a poorly-drained floodplain facies comprised of interbedded deposits of sand-poor and sand-rich beds, green/grey mudstone associated with coal, carbonaceous shale, and minor limestone; and an open-water facies (marine deposits) of sandstones and siltstones with thin limestone [22]. The strata dip to the south at approximately 21°.

Figure 3.
Map showing the extent of the onshore and offshore regional Maritimes Basin extent. The Cumberland Subbasin is also included in the Maritimes Basin, but has been separately highlighted. The three major fault zones are as follows: CCFZ, Cobequid-Chedabucto Fault Zone; CFZ, Cabot Fault Zone; and HFZ, Hollow Fault Zone (modified from [47][48][49][50]). The acronym "PEI" stands for Prince Edward Island.  [24,37]). The stratigraphic column shows the geological formations with their relative ages. The formations, from both the Cumberland Group and Mabou Group, make up the conformable 14.7 km of strata coastline that is UNESCO recognized.

Equipment and methods
To achieve the research objectives of this study, a GPR and RTK DGPS system was used. To assist with GPR interpretation, data gathering was paired with an RTK DGPS to precisely georeference the GPR data. The Joggins Formation outcrop along the shoreline also provides a valuable 2-D view, while the GPR attempts to add the 3 rd dimension. The outcrop study of the Joggins Formation helps to characterize and confirm the sedimentology and the internal architecture of the fluvial outcrop, particularly at smaller scales, which the GPR imaging is unable to resolve.
A Sensors and Software Incorporated pulseEKKO Pro SmartCart GPR system ( Figure 5) was used for this study and supplied by the Dalhousie University Basin and Reservoir Laboratory. The cart is highly durable and has 4-wheels to provide the rapid and continuous collection of data in open areas. The cart is a self-contained system that includes the GPS rover receiver, transmitting and receiving antennae, digital video logging screen, and power supplies for the screen and antennae. The triggering method for the GPR survey is the built-in odometer. The transmitting and receiving antennae were oriented perpendicular to the line profile direction. The 50 MHz antennae with the standard transmitter/receiver configuration and a separation of 1.0 m was used for all lines. It was chosen because the specifications suggested a depth of penetration and resolution sufficient for Joggins Formation imaging. The 50 MHz antennae have a length of 2 m and a nominal spatial resolution length of at most 0.5 m. The step size is 0.1 m, the time window is 400 ns and there are 250 points per trace. The sampling interval is 1600 picoseconds (ps). The transmitter pulsar voltage is 1000 volts. The assumed velocity was 0.100 m/ns, which is between the value for wet clay and dry clay. The imaging of the Joggins Formation is captured to a depth of approximately 300 ns two-way travel time (TWT), corresponding to a depth of roughly 17.0 m. Setup parameters are listed in Table 1. The GPR system was paired with a RTK DGPS to provide a fully georeferenced group of survey lines with positional accuracy of approximately ±2 cm [17]. The survey incorporated a Leica GPS1200+ Series High Performance Global Navigation Satellite System (GNSS) to apply differential corrections and broadcast accurate location data to the rover receiver. The GPS system consisted of a base station and a transmission antenna used to transmit corrections from the base station to the rover receiver in real-time ( Figure 6). The rover receiver was mounted to the midpoint of the GPR cart. The base station was placed over a drilled water well with established surveyed coordinates (UTM Zone 20T; Easting = 387,098.72; Northing = 5,061,126.31; Elevation = 26.453 m) at the rear (water side) of the Joggins Fossil Cliffs Centre. As the GPR cart is pushed along a survey line, the rover receiver acquires GPS coordinates and the wander or drift that is recorded by the base station is subtracted in real-time from the coordinates recorded by the GPR cart. The corrected points are then recorded into the radargram.  Figure 6. Sketch of the global positioning system equipment used to provide a fully georeferenced ground-penetrating radar data set (modified from [17,48]). The base station was assembled over a drilled water well with known coordinates. The transmission antenna was placed adjacent to the base station. The rover was mounted on the GPR SmartCart for providing survey line locational coordinates.
The processing workflow follows three main tasks; the first being the selection of an acceptable GPR data processing workflow; the second being the selection of the appropriate parameters and inputs for each processing step, where required (the dewow filter does not have any inputs, it is simply applied or not applied); and finally, the observation of end results for each processing step and the correction of any issues caused by an incorrect parameter [51]. EKKO_Project™ software was used for editing, processing, and viewing the GPR data. The software was developed by Sensors and Software Incorporated and is a professional "software version that allows for data plotting, editing and full processing routines including spatial and temporal filters, migration, instantaneous attributes, amplitude spectra, CMP velocity analysis and more." All GPR data were post-processed. Processing was performed using iterations paired with descriptions of how and when each processing technique should be applied. Processes were applied individually and in conjunction with other processes until the radargram was sufficient for interpretation.
The GPS data was collected concurrently with the GPR data on the Sensors and Software Digital Video Logger (DVL) and was added to the GPR data using the file that recorded GPS positions at regular trace intervals. The GPS data was stored as the GGA (Global Positioning System Fix Data) format, which is a standard format recognized by the National Marine Electronics Association. The GPS data was converted to UTM coordinates and the step-size re-calculated. Topographic correction of the GPR data along the survey lines was performed using the EKKO_Project™ software. The topographic variation along the survey areas is shown on the inset elevation profile plot in Figure 7a.
Signal saturation correction or dewow is a type of time filter and was applied to each trace for the removal of the initial DC component and low-frequency, slowly decaying "wow" [52,53]. This is caused by the arrival of early waves, dynamic range limitations on instrumentation, and/or inductive coupling effects and becomes superimposed on the high frequency reflections [52,53]. It is typically almost always applied and is usually the first process applied.
To compensate for the spreading and attenuation of the propagating wave front, the Spreading and Exponential Compensation (SEC) gain was used to apply the exponential gain (approximately 1/r2) that compensates for the spreading and attenuation of the propagating wave front. The input parameters necessary for this gain are an attenuation value for the substrate, a beginning value to be added to the exponential gain function and a maximum value for the gain. The average time-amplitude plot for each trace was examined both before and after the application of the gain to ensure it was properly applied as described by Annan [54]. Each line had unique attenuation, start and maximum values. The values for attenuation ranged between 2.08 and 5.31 dB/m with an average of 3.63 dB/m. The start gain value ranged from 0.62 to 1.44 with an average of 0.92. The maximum gain value applied ranged from 32 to 222 with an average of 123.79.  Table 2). The roads are primarily non-linear, so the survey was completed using numerous short straight-line segments. The GPR survey areas are sparsely populated; however, there are an abundance of surficial features, both human-made and natural, that may have varying effects on the quality of the GPR data collected. Many of the surficial features and objects that occur adjacent to or in the vicinity of GPR data collection are shown in Figure 7b. Residential dwellings are generally well-spread out, but are located in certain areas along Hardscrabble Road, Main Street, and Mitchell Street. Along with these dwellings is the associated infrastructure, mainly utility poles and power lines, which do occur in abundance adjacent to the three road surfaces surveyed. The infrastructure also includes steel guard rails with wooden posts along the road areas that closely border the cliff edge. Areas of vegetation (e.g., grass, plants, and trees) are also located along the flanks of the roads where the GPR surveys were completed on. Due to the variation in surficial features along the GPR survey lines, it may be possible to look for subtle changes in the reflection profiles and correlate the response to a particular feature.

Processed radargrams
The completion of the GPR survey at Joggins, Nova Scotia resulted in 42 unique radargrams across three road surfaces (Hardscrabble Road, Main Street, and Mitchell Street) and a grassy area adjacent to Main Street. The 42 processed radargrams are displayed in Figure 9. All radargrams have an initial high amplitude, thick, horizontally continuous reflector. It always occurs at the top of the radargram and follows the topography of the individual survey line.
In addition, all radargrams display a second high amplitude, thick, primarily horizontally continuous reflector that is always below the primary top reflector. It too will generally follow the topography of the individual survey line, but can be discontinuous. Numerous radargrams contain high amplitude, sharp, concave downwards reflectors that have a consistent shape and are either overlapping or have a consistent spacing. In general, coherent reflectors are absent below approximately 6 to 8 m depth. The resulting reflectors that are visible in the radargrams are summarized in the following radar reflectors section and in Figure 10. In total, 13 radar reflectors are described and interpreted.

Radar reflectors
Selected results from four areas surveyed across the Joggins Formation are presented here. The variability in surficial features along the various surveys results in reflectors that can be correlated nicely to these features. A total of thirteen unique radar reflectors have been identified from the GPR data and are briefly summarized ( Figure 10). RR1 is a high amplitude, thick reflection and is the first signal measured by the receiver and occurs in all 42 collected radargrams. It is continuous for the complete length of each radargram and is followed by a low amplitude signal. It follows the natural topography of the GPR survey line. There are no other features that occur with this reflector. RR2 is a high amplitude, thick reflection and is the second signal measured by the receiver and occurs in all 42 collected radargrams. It is typically continuous for the complete length of each radargram, although it can be discontinuous in several lines. It is followed by a low amplitude signal. It follows the natural topography of the GPR survey line. This reflector is affected by the subsurface and surficial features. Reflectors RR3 is characterized by a thin, high amplitude reflection, followed by a lower amplitude signal. This reflector occurs at approximately 8 m depth in the radargrams it occurs in, which is consistent with the approximate depth of the overburden in the area. The RR4 reflector is common throughout the radargrams. It is characterized by regularly spaced; high amplitude followed by low amplitude concave downwards reflectors. They have a consistent shape and are sharply outlined.
The RR5 reflector is also common throughout the radargrams. It is characterized by irregularly spaced; high amplitude followed by low amplitude concave downwards reflectors. They have a les consistent shape and are not as sharply outlined as the RR4 reflector. reflector RR6 is not common throughout the radargrams and is characterized by a break in the second thick, high amplitude reflector. The RR7 reflector is characterized by high amplitude, repeating, and parallel reflectors that reverberate throughout the radargram. The RR8 reflector is reflection-free. There are no coherent reflectors observed in the radargrams. RR9 is a thick, high amplitude, concave downwards reflection that only appears to be visible in a couple of radargrams. It is associated with the high amplitude, thick, continuous reflector (RR2). RR10 is a high amplitude, thick, and discontinuous reflector that occurs below RR2. It displays an undulating profile and is only visible in a single radargram. RR11 is a continuous, upright, high/low amplitude reflector that mimics ra teepee shape. It is very-well defined and sharp. It occurs in only five radargrams. RR12 is a vertical, mottled high and low amplitude reflector that is only visible in a single radargram. The features are not continuous throughout the whole radargram. RR13 is a small, near-surface, high amplitude reflector. It is concave downwards and sharp with a consistent shape. They are associated with RR2.

Discussion
RR1 is defined as the direct air wave. This high amplitude reflection is the first signal measured by the receiver and occurs in all 42 collected radargrams. It is continuous for the complete length of each radargram and is followed by a low amplitude signal. Typical examples of this radar reflector can be seen in all radargrams, including radargrams 31 and 37 ( Figure 11). Since the radargrams have all been georeferenced, the direct air wave follows the topography. RR2 is defined as the direct ground wave. This high amplitude reflection is the second signal measured after the direct air wave and is also present in all 42 radargrams. The majority of the radargrams display a continuous direct ground wave, except for lines 27, 39, and 45 in which they are discontinuous. Typical examples of this radar reflector can be seen in all radargrams, including radargrams 31 and 43 ( Figure 12). The undulating nature of the ground wave highlights the small topographic changes that occur over the lengths of each GPR line segment. RR3 is interpreted to represent the angular unconformity contact between the Carboniferous-aged Joggins Formation and the overlying Quaternary-aged glacial till and soil cover. The sharpness of the contact as well as the depth it is occurring at (6 to 8 m) are the main reasons for this interpretation. Curiously though, this reflector is not present in all radargrams, even though all survey lines were completed on top of overburden overlying the Joggins Formation. Examples of this radar reflector can be seen in several radargrams, including radargrams 09 and 13 ( Figure 13). The RR4 reflector correlates to utility poles that are erected at certain locations adjacent to the road surfaces. Several radargrams were recorded along road surfaces that have regularly spaced wooden utility poles located less than 10 m from the centerline of the road. The data from these lines show regularly spaced, clear, and sharp, concave downwards reflectors whose locations match those of the utility poles along the survey lines. Typical examples of this radar reflector can be seen in several radargrams, including radargrams 18 and 23 ( Figure 14). The RR5 reflector is also composed of concave downwards reflectors. However, these reflectors are noticeably different when compared to the utility poles described by RR4. Through careful examination of the radargrams, it was determined that RR5 represent trees. In contrast to the regularly spaced utility poles giving regular and clear diffraction hyperbolae, trees on the other hand produce hyperbolae that are randomly occurring and are overlapping. This theory was tested by viewing the GPR lines that did not have power line infrastructure located next to the road surface, but did have abundant, randomly occurring trees. Representative examples of this radar reflector can be seen in radargrams 10 and 32 ( Figure 15). The RR6 reflector is rarely observed in the radargrams. The survey lines where this reflector becomes discontinuous are also the areas where the GPR has passed beneath a power line. It can therefore be surmised that overhead power lines can cause a brief break in the direct ground wave. Typical examples of this radar reflector can be seen in radargrams 27 and 39 ( Figure 16).  We interpret the RR8 reflector as areas where the GPR signal has been attenuated. This typically signifies either a lithological unit that is massive and homogenous, the presence of dissolved minerals in groundwater with highly conductive properties, and/or the presence of clay-rich sediments (e.g., [56,57]). In this study, it was determined that the attenuation was caused by the overlying clay-rich glacial till and soil. This reflector makes up the majority of all radargrams and is widespread among all 42 radargrams, indicating that at least locally, the clay-rich glacial till and soil are probably present. Typical examples of the attenuated signal can be seen in radargrams 19 and 20 ( Figure 18).  RR10 is interpreted to represent the undulating angular unconformity contact between the Carboniferous-aged Joggins Formation and the overlying Quaternary-aged glacial till and soil cover. The sharpness of the contact as well as the depth it is occurring at (6 to 8 m) are the main reasons for this interpretation. The lone example of this radar reflector can be seen in radargram 30 ( Figure 20).  RR12 is interpreted to be two instances where data skips occurred (not collected at a trace location). These two vertical features are not related to any subsurface features. The sole example of this radar reflector can be seen in radargram 25 ( Figure 22). Despite these two instances of non-signal, there are noticeable repeating hyperbolae occurring in the background, which correlates to the wooden utility pole at position 75 m near the end of the line. RR13 is interpreted to represent the locations of boulders and are associated with the RR2 radar reflector (direct ground wave). Since the area does contain several meters of glacial till, it would be reasonable to assume that larger rocks are present beneath the areas traversed with the GPR. Typical examples of this radar reflector can be seen in radargrams 09 and 10 ( Figure 23).
Performing a GPR survey as it was done over the Joggins Formation presented a number of challenges. In the 42 radargrams that were gathered over the study area, it does not appear that any contain reflections that would be considered those of the Joggins Formation. The most probable culprit for the failure in imaging the Joggins Formation strata is the thick clay-rich soil and glacial till overburden that would have greatly attenuated (radar reflector 8) the transmitted energy in the subsurface. Another probable culprit is the Joggins Formation itself. It is well known from viewing the outcrop on the intertidal area that the strata dip at a constant 21° and are highly variable with respect to both lithology and thickness. Sedimentary beds of a certain thickness would not be visible since they are below the resolution of the antennae used. Furthermore, the beds are composed of a wide range of lithologies, from clay-sized particles up to gravel-sized particles; thus, the individual beds themselves could be contributing to attenuation as well. Power lines run along the edge of the roads that were surveyed, which may result in some problems. At certain sections of the survey, the GPR was near the cliff edge. This could translate into some edge effects in the radargrams.
The GPR data are contaminated to varying degrees by above-ground objects. Several lines were recorded along road surfaces that have regularly spaced wooden utility poles located less than 10 m from the centerline of the road. The data from these lines show regularly spaced and crisp/clear In contrast to the regularly spaced utility poles giving regular and clear diffraction hyperbolae, trees on the other hand also cause a similar phenomenon in the radargrams, except that the hyperbolae occur randomly and are overlapping. This theory was tested by viewing the GPR lines that did not have power line infrastructure located next to the road surface, but did have abundant, randomly occurring trees. Typical examples are shown in radar reflector 5, which displays abundant, irregularly spaced and overlapping hyperbolae. Similar diffraction hyperbolae occur in lines 9 to 13 and 31 to 35 on Hardscrabble Road, and Line 43 on Main Street.
In a few instances, a power line will cross over Hardscrabble Road or Main Street. When this occurs, a noticeable feature can be observed in the radargrams (radar reflector 6). The clear and crisp reflectors that are typical with the utility poles become more chaotic. This would infer that the power lines passing overhead do influence the GPR signals.

Conclusions
In this study, we have utilized GPR in an attempt to image the internal geometry and architecture of the Joggins Formation and thus, aid with extending the 2D outcrop into 3D by way of modeling the larger scale features. Through iterations of GPR processing techniques and examination of the resulting radargrams, it can be concluded that the survey was unsuccessful in showing Joggins Formation structures and internal architecture. This is likely the result of a combination of factors, with the dominant one being the thick, highly conductive nature of the clay-rich glacial till/overburden. Unfortunately, it was determined that numerous difficulties relating to both the Joggins Formation itself and the area over which the GPR data was collected made any imaging of the Joggins Formation non-occurrence. The imaging issues that are directly related to the Joggins Formation most likely include (1) dipping beds cause increased refraction, unlike a horizontal or nearly horizontal bed, (2) the scale of the beds is too fine for the GPR configuration used for individual beds; at best, it could only image the thicker beds, and (3) the abundant jointing and fracturing visible in the outcrop exposure probably permeates throughout the subsurface, thereby compounding imaging problems. It is possible that lower frequency antennae could have provided a depth of penetration sufficient to image the Joggins Formation; however, the resolution would have been affected and the likelihood of being able to interpret the dipping strata would have been compromised.
A variety of overburden and surface/subsurface objects may also affect the GPR data collection. Some of the most likely issues include (1) the soil and glacial till making up the overburden layer are clay-rich, thereby leading to signal attenuation, (2) the compacted road surface also has clay materials, in addition to other materials which would also enhance signal attenuation, (3) metallic/wooden objects either exposed at the surface or buried will create additional artifacts in the radargrams. These objects include galvanized steel guard rails, overhead power lines and their associated infrastructure, traffic signage, and cars parked in driveways or passing the GPR equipment while collecting data. Other potential issues that may hamper efforts include the uneven terrain the data was collected on and whether there are GPR data collection issues near the edge of a vertical cliff face. The effects of surface objects, such as trees and utility lines are well-documented as being the culprits of many diffraction hyperbolae seen in the radargrams.
Despite the lack of subsurface imaging from the Joggins Formation, significant insight was gained as to the limitations of the GPR application in this type of environment. There was also some moderate success in imaging the angular unconformity, which was apparent in several radargrams. Perhaps the imaging of the unconformity was related to variations in overburden thickness and/or clay content. Although the primary objective of imaging the Joggins Formation was unsuccessful, it nevertheless increased our knowledge concerning the true impact of clay-rich overburden sediments on bedrock imaging, in addition to the impacts that surface features can cause on GPR data collection. The knowledge gained from this study can be utilized in future GPR surveys, particularly with respect to geoforensic studies where the study areas contain similar surface infrastructure (e.g., utility poles, power lines, etc.). This study also demonstrates the usefulness of shallow subsurface GPR for geohazard assessments. It is important to reiterate that this was the site of extensive past coal mining efforts with well-developed surface infrastructure. Therefore, the area probably contains an abundance of erratic metal objects, large boulders, etc., that would have an influence on the generated radargrams.