Case Histories of GPR for Animal Burrows Mapping and Geometry

Abstract

Ground-penetrating radar (GPR) has a wide range of applications, from geologic mapping to concrete inspection. A recently emerging GPR application is deployment in biological investigations as a non-invasive technique. Geophysical mapping of features such as tree roots and turtle burrows has proved valuable for the understanding of these subsurface systems for ecological, environmental, or engineering purposes. Four case histories of GPR investigations pertaining to animal burrows are described: cutter ants in Brazil, groundhogs in Michigan, and groundhogs, and burrowing bees in Pennsylvania. Cutter ants (Atta spp.) in Amazonian Brazil are known to construct burrows of nearly the same dimensions as groundhogs as they excavate galleries up to 7 m deep for leaf storage. Cutter ant burrows are hazardous to heavy equipment and may also cause loss of mud circulation during rotary drilling. Groundhogs (Marmota monax), found throughout the United States, cause unseen hazards, particularly for equestrian facilities where a sudden collapse can cause severe injuries to both horse and rider. Burrowing bees (Colletes inaequalis) are common in the northeastern United States. The size of the bee burrows is significantly smaller than that of the cutter ants and the groundhogs. The data for these surveys were collected over a twenty-year span, crossing several generations of survey equipment and processing techniques. Together, these four case histories highlight the historic and current capabilities of GPR systems applied to mapping subsurface burrow systems. These examples demonstrate the important impact near surface heterogeneities have in altering ecological, environmental, or engineering systems and the utility of GPR for mapping such heterogeneities.

Introduction

Ground-penetrating radar (GPR) has been used for a wide variety of geological applications over the last several decades. This geophysical method has been successfully applied in mapping landslides (Sherrod et al., 2014; Sass et al., 2008; Barnhardt, 2000), deciphering glacial history (Blewett et al., 2014; Choi et al., 2014), concrete inspection (Baryshnikov et al., 2014; Varela-Ortiz et al., 2013; Zanzi and Arosio, 2013; Diamanti et al., 2017), and identifying zones of contamination (Cassidy, 2007; Porsani et al., 2004; Sauck et al., 1998), among many other applications. Additionally, the versatility of GPR techniques is transferable to archaeological exploration because of the ability to detect slight changes in the subsurface electrical properties making the method ideal for locating artifacts, foundations, and previous excavations. Archaeological investigations around the world have benefited from GPR analysis (Gaff et al., 2013; Porsani et al., 2010; Sarris et al., 2007; Conyers and Goodman, 1997). However, often GPR results are inherently ambiguous. For example, field sites can have multiple subsurface sources of GPR reflections that are not the intended survey target. Animal burrows and tree roots, for example, are frequently a source of noise within GPR profiles at archaeological sites (Gaff et al., 2013; Porsani et al., 2010; Sensors & Software Inc., 2010; Rodrigues, 2009).

Noise for one study, however, can be the target of another investigation. Biologists and botanists are applying GPR techniques to define burrow and root systems. The GPR signatures created by tree roots, for example, have been used to develop estimates of biomass (Simms et al., 2017; Guo et al., 2013). Burrow mapping has been accomplished for rabbits (Stott, 1996), crabs and foxes (Grimes et al., 2011), moles and groundhogs (Grimes et al., 2011), gopher tortoises (Kinlaw and Grasmuek, 2012; Martin et al., 2011), wombats (Swinbourne et al., 2014; Swinbourne et al., 2015; Swinbourne et al., 2016), and badgers (Nichol et al., 2003). The purpose behind these burrow investigations varies from biological interest, environmental, and civil engineering applications. Ichnological research and the field of zoogeomorphology are additional areas that have recently applied GPR to burrow mapping (Buynevich et al., 2014; Hembree et al., 2017; Kopcznski et al., 2017). There is significant interest in burrow complexity for identifying and characterizing trace fossils (Simpson et al., 2010; Hembree and Hasiotis, 2008; Hasiotis et al., 2007; Hickman, 1990). Differentiating between classes of animals is of particular importance in this field. Mammalian burrows are typically complex, with multiple entrances and tunnels around a sleeping chamber, whereas reptilian burrows are expected to be simple, with one entrance and a chamber (Storm et al., 2010; Martin et al., 2008). Insect burrow morphology varies widely, with the large range of sizes of insects, the many life cycle stages of these animals, and occasional colonial nature all playing a part in the variations (Hembree and Hasiotis, 2008; Smith and Hasiotis, 2008).

Environmental purposes focus on understanding preferential flow within the unsaturated zone and the impact biological burrows have in forming preferential flow pathways affecting fluid fate and transport. Connected burrows have been shown to reduce fluid residence time within the unsaturated zone by providing conduits for fluid to quickly enter the subsurface (Chappell, 2010; Golab et al., 2017; Klaus and Zehe, 2010; Sander and Gerke, 2009; Zehe et al., 2010), provide a primary path for aquifer recharge (Stahl et al., 2014), and result in preferential pathways for contaminant transport (Dadfar et al., 2010). Alternatively, burrow dead ends have been found to intercept flow and slow or reduce groundwater influx (Carol et al., 2011; Chappell, 2010; Wang et al., 1996). Engineering studies of animal burrows include highway road cut stability or dam and levee safety objectives. For example, Chlaib et al. (2014), Di Prinzio et al. (2010), and Barner (2001), provide comprehensive summaries of the importance of mapping burrows in levees. GPR is not the only geophysical method that has been employed to investigate burrow structures. For example, Butler and Roper (1994) employed resistivity and magnetometer techniques to map badger burrows in England. The resistivity results, however, were not as clear as the GPR images and the magnetometer results did not show any conclusive evidence of the burrow structure. Understanding the preferential flow impacts for environmental purposes and the hazards associated with burrows near engineered structures is critical, but the pool of literature for burrow investigations, mapping, and characterization is limited.

We present four studies which demonstrate the use of GPR in relation to mapping of animal burrows across a wide range of locations, geologic environments, and eras of equipment. These include: cutter ants in Brazil, groundhogs in Michigan, and groundhogs and burrowing bees in Pennsylvania (Fig. 1). At the first site, we demonstrate the application of GPR for imaging the burrows of cutter ants at the Fazenda Aruanã Brazil nut tree plantation where the insects are causing damage to crop productivity. The second site is an investigation of groundhog burrows in relation to equestrian hazards at an indoor riding arena in southwestern Michigan. Two separate surveys in eastern Pennsylvania were performed to define the complexity of mammalian burrows and insect burrows for comparison to burrows in the rock record. Throughout this manuscript, the terms tunnel, burrow, and burrow system or burrow complex are used to represent a linear portion of the burrow, the entirety of the dwelling of an animal, and the interconnected tunnels that comprise the dwellings of animals respectively. These four case histories serve as a representative sample of the use of GPR in search of burrow systems.

Figure 1.

Map of the four field sites (adapted from Google Earth). The coordinates for each site are: Site 1) 3.04°S, 58.77°W; Site 2) 42.78°N, 85.50°; Site 3) 40.32°N, 75.72°W; and Site 4) 40.50°N, 75.77°W.

Theory

GPR is an electromagnetic geophysical method that sends an electromagnetic (EM) pulse into the subsurface and measures the travel time and amplitude of the signal reflection when it returns to the antenna. Reflections of the EM pulse are caused by subsurface changes in the electromagnetic properties. Large changes in subsurface properties result in strong amplitude reflections. Thus, the transition between sediment types can be imaged. Likewise, voids in the subsurface represent a very drastic change in electrical properties and typically generate strong reflections.

Subsurface electrical conductivity is closely linked to both quality of signal and depth of penetration. Skin depth is defined as the depth at which the amplitude of the EM signal attenuates to 1/e, or approximately one third of the original amplitude, and may be calculated using:

$$d_s=503\sqrt{\frac{\rho }{f}}$$ (1)

where d_s is skin depth in meters, ρ is resistivity in ohmmeters, and f is the frequency in Hertz (Reynolds, 2011). This relationship demonstrates that the depth of penetration of the GPR signal is directly proportional to the square root of the resistivity of the Earth and inversely proportional to the square root of the GPR antenna frequency. To image the subsurface at great depths, a low frequency antenna is required. Likewise, high resolution imaging of the shallow subsurface requires a high frequency antenna to provide the best results. A subsurface with low electrical resistivity will cause the signal to attenuate more rapidly with depth than a subsurface with high electrical resistivity. Hence surveys performed on sandy soils produce better results than surveys performed on clay-rich soils (at least in temperate zones).

Dielectric permittivity also has a significant impact upon the GPR signal. Van Dam (2014) provides a thorough evaluation of the importance of this value and its use in GPR applications. The value of the subsurface relative permittivity determines the velocity of the EM pulse as it travels through the subsurface. Estimates of this property may be used to determine the approximate depth of reflectors according to the following formula:

$$d=\frac{ct}{2\sqrt{{ϵ}_r}}$$ (2)

where d is the depth, c is the speed of light, t is the two-way travel time of the reflected EM pulse, and ϵ_r is the relative permittivity of the subsurface (Reynolds, 2011). Dry, sandy soils have a relative permittivity range of 3 to 6 whereas moist soils can have a relative permittivity over 20 (Reynolds, 2011). The water table was below the imaged profile in all cases presented in this work. The relative permittivity can be estimated through a ground truth measurement of an interface observed in an excavated profile, analysis of common midpoint data, direct relative permittivity measurements via a dielectric probe or time domain reflectometry, or matching diffraction hyperbolas within the GPR profiles to move-out calculations for different relative permittivities. The latter of these methods was used to estimate depth in the case histories presented in this work.

Case History 1

Site Description

The first field site is a Brazil nut tree plantation, Fazenda (Farm or Ranch) Aruanã, located north of the Amazon River between Manaus and Itacoatiara, Brazil (Fig. 1). The plantation owners began cultivating the area in the mid-nineteen eighties, but by 1994, no nuts had been produced and many of the trees were not healthy. Scientists were hired to pinpoint the source of the decreased productivity at the plantation and the unhealthy trees. The causes were determined to be vegetation, soil properties, and cutter ants. The main problem, identified and addressed by the biological investigators was that the tropical rainforest vegetation had been cleared in preparation for a monoculture of Brazil nut trees planted at a spacing of 10 m. In so doing, the owners had unwittingly eliminated the habitat of the sole pollinator of the cash crop, a small wasp that lives in the underbrush in areas around Brazil nut trees. However, beyond this habitat dilemma, the production of the Brazil nut trees at this location still suffered and the cause was still in question. Even though the soils in this rainforest are very productive, certain locations within the plantation seemed to result in healthier growth than other locations. Additionally, the work of leaf cutter ants had killed several acres of trees as they harvested leaves and carried them back to their burrows.

The cutter ants have lengths of up to 1.0 to 1.5 cm. This species carries the leaves to an extensive burrow system to cultivate fungi. In addition to killing vegetation, cutter ants create major problems for people in the areas surrounding their burrow systems. Biologists have traced galleries to depths of up to 7 m. Such extensive burrow systems represent a significant hazard for heavy equipment because the excavated burrow system could collapse under the equipment weight causing damage to the device and operator. In addition, mud rotary drillers have been known to lose as much as 1.9 m³ (500 gallons) of mud into these types of burrow complexes adjacent to the operations causing loss of mud circulation and potential damage to the drill rig.

The area of interest at Fazenda Aruanã is 50 km², which includes a series of experimental Brazil nut tree plantation test plots. An inactive colony at the plantation (Aruanã Site) and an active colony near the plantation (Saúva Site) were identified by surface observations. Ant “hills” were 2–3 m in diameter and 20–30 cm high. Most of the holes were 5–15 cm in diameter. Trails leading to the burrow entrance of active colonies were cleared by the ants to make leaf transportation to their burrow more efficient. These trails were 10–15 cm wide and radiated in all directions from the holes. Trees at the Aruanã Site were reportedly doing poorly in the sandy soils, formed from shallow sandstone below the residual soils in the west, and well in the clayey soils going up a slope to the east. An extinct cutter ant colony was visible about halfway up the slope. The Saúva Site was near the Brazil nut tree test plot area, on a ridge or plateau top within clayey terrain. The active colony was in the middle of several acres of brush, where the surrounding Brazil nut trees had all been killed. GPR investigations were performed in September 1994 to characterize the soil and image the cutter ant burrows at Fazenda Aruanã.

Data Collection and Processing

GPR surveys were performed in two locations: a long profile perpendicular to the transition between trees that were growing well and trees that were growing poorly (Aruanã Site) and several short profiles over the active cutter ant colony (Saúva Site). A GSSI SIR 10 system with a 500 MHz antenna and a monostatic custom-built 100 MHz antenna using the GSSI 769DA transceiver antenna driver card was used to acquire the data. The profile at the Aruanã Site, over the transition zone, was performed with the 100 MHz antenna between one of the rows of Brazil nut trees. The antenna was pulled by a motorized vehicle to generate a continuous 250 m profile across the transition zone. Multiple scan lengths and gains settings were used in repeat runs to image the inactive cutter ant burrow. Following this profile, investigations over an active cutter ant colony (Saúva Site) were performed. Six profiles each 13 m long and spaced at 1m intervals were surveyed with the 100 MHz antenna. The last four of these profiles were resurveyed with the 500 MHz antenna. The scan length was set to 370 ns for the 100 MHz survey of the Aruanã Site, whereas the settings for the Saúva Site were 200 ns for the 100 MHz antenna and 80 ns for the 500 MHz antenna. Data processing was performed in Reflexw (http://www.sandmeier-geo.de/index.html) and included distance normalization, stacking of 7 scans, background removal, bandpass filtering, and range gain as appropriate for each profile.

Results

Tunnels in the subsurface are expected to produce either a flat reflection of increased amplitude in the GPR profile (for surveys run parallel to the tunnel) or a hyperbolic reflection of increased amplitude in the GPR profile (for surveys run perpendicular to the tunnel). The inactive cutter ant burrow system was surveyed at the Fazenda Aruanã Site. The survey over the transition zone, clearly shows the difference between the GPR response in sandy soils and clay-rich soils (Fig. 2(A)); displaying the full profile length). An immediate decrease in the amplitude of the reflections occurs at 135 m east along the profile. Elevation measurements were not taken in the field, and the profile is not corrected for topography, however a slight increase in slope was noted at the field site halfway through the profile. A relative permittivity of 5 was assumed throughout the entire profile based upon the sandy soil and matching diffraction hyperbolas within the GPR profiles to move-out calculations for different relative permittivities. The inactive cutter ant burrow system is located within the clay-rich section of the profile and demarked with a rectangle in Fig. 2(A). The hyperbolic reflections from this burrow system are apparent in this attenuated zone of GPR reflections at a distance of 180–240 m east within the profile (see Fig. 2 (B) which provides a display zoom for this distance range with the hyperbolic burrow reflections highlighted with dashed lines).

Figure 2.

Cutter ant burrow GPR image with the 100 MHz antenna at the Aruanã field site showing: A) the transition from sandy to clay-rich soils (vertical line) and inactive cutter ant burrows (rectangle); and B) zoomed-in view of the inactive cutter ant burrows from 180E to 240E. Hyperbolic reflections related to the burrow system are traced in bold. The relative permittivity was estimated at 5 for all profiles in this figure, indicating a velocity of 0.13 m/ns.

The active colony of cutter ants is located in the Saúva Site, adjacent to Fazenda Aruanã. Figure 3 highlights the depth vs. resolution trade off by showing the lower resolution in Fig. 3(A) from the 100 MHz antenna and the reflection detail evident from the 500 MHz antenna (Fig. 3(B)) due to the increased resolution. The 500 MHz data clearly illustrates the complexity of the active burrow system.

Figure 3.

Cutter ant burrow GPR image at the Saúva field site, adjacent to Fazenda Aruanã, depicting the resolution of: A) 100 MHz antenna on an active cutter ant burrow (circled); and B) 500 MHz antenna on an active cutter ant burrow (circled). Note the increased resolution of the burrow system with the higher frequency antenna. The relative permittivity is estimated at 5, indicating a velocity of 0.13 m/ns.

Discussion

A shallow and slightly undulating sandstone bedrock reflection is observed at the Aruanã Site at a depth of 6 m in the western half of the profile, west of the transition to the clay-rich soil zone (Fig. 2(A)). This reflection is lost in the eastern half of the profile (east-west boundary shown with vertical black line), which corresponds to the increase in the subsurface clay content and an increase in the slope of the land surface. The transition between sandy to clay-rich soils represents a transition from low relative permittivity within the sand to significantly higher relative permittivity from the clay. A relative permittivity of 5 was used as the base value to calculate the depth scale throughout the entire length of the profile. Thus, the reflections on the east side of the profile are plotted at a greater estimated depth than is real, indicating that the cutter ant burrows imaged in this profile extending to an estimated depth of 6 m are in fact not quite that deep. However, the shape of the burrows as well as the expansive nature of the subterranean system is evident, with hyperbolic reflections from the burrow system dominating the GPR response from 186 m to 236 m along the profile (Fig. 2(B)), for a total distance of 50 m. Field observations conclude the major entry into the tunnel system is centered at 189 m, with several smaller tunnels extending outward from that to the east. This location represents an inactive or extinct burrow system with no ants actively maintaining the structure.

The results from the Saúva Site clearly display the resolution benefits of using a high frequency antenna (Fig. 3). The 100 MHz antenna provides a deeper subsurface image, but as the burrows are confined to the upper 4 m of the subsurface, this depth of penetration is unnecessary. The results from the 500 MHz antenna (Fig. 3(B)) show numerous intersecting hyperbolic reflections from 6 to 10 m northeast along the profile, representing tunnels created by the cutter ants in their burrow system. Comparison of these burrow reflections to the ones of the inactive burrow system shows that the active burrow system has approximately the same GPR reflection dimensions for the major entry into the tunnel system. The active system has more interference of the hyperbolic reflections that may indicate more small tunnels active at the entrance. The inactive system was surveyed over a distance of nearly 5 m. A longer profile over the active system may have shown additional tunnels in the surrounding area, or the size of the active system may not yet have reached its maximum potential.

Case History 2

Site Description

The soils in Michigan’s Lower Peninsula are formed in glacial deposits. Caledonia, Michigan is known for glacial soils of sands and gravels derived from glacial outwash and moraines. At Oosting Riding Arena near Caledonia in 2008, a horse with rider stepped on and broke through a groundhog burrow in the indoor arena. No injury to the horse or the rider occurred, but the owners of the facility were concerned. Such interaction with groundhog burrows can cause horses who misstep to potentially break a leg. Geophysicists were called to the site to investigate the extent of the burrow system at this site in October 2008. The burrow system exposed by the above incident was backfilled by the owners prior to the geophysical investigations.

Data Collection and Processing

A GSSI SIR 10A+ GPR system with 500 MHz antenna was used to survey the groundhog burrow system. The arena dimensions are 18 m north-south and 55 m east-west. GPR data were collected on a series of 18 m south to north survey lines with a line spacing of 0.6 m from 0E to 14E. An anomalous GPR response zone was detected during the field data collection. Additional lines were surveyed west to east at this location along lines 8.5N to 12.2N, starting at 0E and extending to 18E with a line spacing of 0.6 m. A range of 40 ns was used for data collection, with unidirectional surveying. Data processing in Reflexw included distance normalization, background removal, bandpass filtering, a subtracting average function, and a range gain function as appropriate for each profile.

Results

The data were inspected for anomalous response zones of hyperbolic reflections and small groupings of increased amplitude which are interpreted as due to burrows. The depth of penetration is estimated at 2.6 m with an assumed relative permittivity of 5 based upon the sandy soil and matching diffraction hyperbolas within the GPR profiles to move-out calculations for different relative permittivities. The north-south lines show hyperbolic reflections between 4.5N and 12N. A series of discontinuous, intersecting hyperbolic reflections was noted along lines 8.5E and 10.5N (Fig. 4). The survey data collection procedure provided a grid which could be used to produce a three-dimensional image, but the wide spacing between lines gave poor resolution for such processing techniques.

Figure 4.

Groundhog burrow GPR image at the Caledonia field site depicting the multiple hyperbolic responses associated with the backfilled burrow in: A) line 8.5E; and B) line 10.5N. The used antenna was 500 MHz and the relative permittivity is estimated at 5, indicating a velocity of 0.13 m/ns.

Although true three-dimensional imaging of the GPR reflections requires shorter line spacing, the pick function within Reflexw or similar GPR processing programs can be used to manually map the burrow locations across the site. User defined location markers are inserted to map subsurface features, in this case the extent of the groundhog burrows. These picks were chosen based upon observation of bright hyperbolic reflections, bright spots, and correlation to adjacent and perpendicular lines. The three-dimensional space is represented for the surveyed area from (0E, 0N) to (14E, 18N) in map view (Fig. 5(A)) and in cube form (Fig. 5(B)).

Figure 5.

A) Map view; and B) three-dimensional view of the manual picks chosen for the identification of the groundhog burrow system through hyperbolic reflections in the data from Caledonia. Manual pick choices from the north-south profiles are shown in black whereas pick choices from the east-west profiles are shown in dark gray. A light gray shading is used to show the interpreted axis of the burrow. The used antenna was 500 MHz and the relative permittivity is estimated at 9, indicating a velocity of 0.10 m/ns.

Discussion

Due to the nature of the glacial soils in this area of Michigan, it is necessary to view adjacent lines to verify burrow features with the use of GPR. Glacial sediments are often poorly sorted. As a result, glacial erratics need to be eliminated as causes of large hyperbolic reflections that could be mistaken for burrows. Confirming the continuity of these hyperbolic reflections across several survey lines as well as in surveys of lines perpendicular to the anomalous reflections is a necessary step in data analysis and crucial to eliminate glacial erratics as a causal mechanism and false positive for animal burrows. The location of the main section of the burrow, as known from the riding incident, is clear from both the GPR profile lines with the high density of discontinuous hyperbolic reflections along these trends (shown within the dashed oval on Fig. 4). Adjacent profiles show similar results in the zone from 6 to 12 m east and 4 to 12 m north. This verifies that the source of the anomalies is a feature of clear lateral extent that is continuous between perpendicular profiles, thus indicating burrows (Fig. 4) and not discrete glacial erratic deposits.

Furthermore, the map of the manual pick choices (Fig. 5(A)) provides a clearer overall view of the burrow locations at the site by visually displaying linear connections between the hyperbolic reflections displayed. Even more informative is the three-dimensional computer visualization of the data which includes burrow depth imaging. The cube display shown in Fig. 5(B) represents a single view of this three-dimensional imaging technique. The burrow reflections at this location are all coming from an approximate depth of 0.5 to 1.5 m below the ground surface. The three-dimensional data processing provides another detailed step toward accurate mapping of the subsurface burrow features. If additional east-west survey lines had been performed, then the interconnection of the hyperbolic anomalies in the southwestern quarter of the survey may have been better resolved; however, these southwestern reflections were lower amplitude and may indicate burrow abandonment. If the tunnels are not maintained or active, then roof and wall collapse may progress and thus leave less of a contrast for the GPR signal in the subsurface. In the current analysis (Fig. 5), sections of the survey area show multiple parallel hyperbolic reflections that are likely tunnels in the groundhog burrow system.

Case History 3

Site Description

The other groundhog site is Rodale Institute located in Kutztown, PA, known in central eastern Pennsylvania for its organic farming, which contain well drained silty loam soils (Ackerman, 1970). An organic upper unit over sandy soil with small amounts of clay is present in the subsurface. This site was chosen not for the hazards that burrows represent, but as a paleontological project in the significance of burrows in the rock record. Groundhogs are an appropriate representative mammal for this project, and the Rodale Institute has an ample supply of such burrows. The subsurface groundhog burrows were mapped with geophysical methods on the south boundary by the east orchard at the Institute (Fig. 1) in December 2011.

Data Collection and Processing

A GSSI SIR 3000 GPR system with both 400 MHz and 900 MHz antennas was used to survey the burrow location at Rodale Institute. GPR profiles were surveyed in both the north-south and the east-west direction across the survey grid. The initial survey, performed with the 400 MHz antenna, was 7 m east-west by 6 m north-south with a line spacing of 0.25 m and a range of 45 ns. This survey covered both an active burrow and an abandoned burrow. The 900 MHz antenna data were obtained on a subsequent day of field work with specific interest in the entrance area of the active groundhog burrow. The purpose of this higher frequency survey was to focus on the intersection of two active tunnels identified in the northernmost section of the 400 MHz survey. This survey covered an area 5 m east-west by 4 m north-south with an origin at (0E, 3N) within the previous 400 MHz survey. These 900 MHz data were collected with a line spacing of 0.1 m and a range of 30 ns. Bidirectional surveying was used to decrease field survey time for both frequencies of antenna. An endoscopic probe was used to view the tunnels from the subsurface to confirm GPR interpretations. Data processing in Reflexw included distance normalization, background removal, bandpass filtering, and range gain as appropriate for each profile.

Results

The data were searched for hyperbolic GPR reflections that could correlate to the structure of the groundhog burrows at this site. The pick function of Reflexw was used to manually identify features relating to the burrows. As for the previous case history, these picks were chosen based upon observation of bright hyperbolic reflections, bright linear reflections, and correlation to adjacent and perpendicular lines. The depth of penetration is estimated at approximately 2.2 m with the 400 MHz antenna and 1.5 m with the 900 MHz antenna using an assumed relative permittivity of 9, estimated based on the soil type at this site and matching diffraction hyperbolas within the GPR profiles to move-out calculations for different relative permittivities. A GPR profile over the main entrance shaft to the burrow and the three-dimensional image of the entrance shaft is plotted with the 400 MHz antenna (Figs. 6(A) and (C)) and the 900 MHz antenna (Figs. 6 (B) and (D)). The interpreted burrow picks are presented in map view and cube form in Figs. 7 and 8, for 400 and 900 MHz survey data, respectively. Both abandoned and active burrows were identified along the north-south profiles (Fig. 7). The entrance shaft, tunnel, ramp, and chamber of this burrow are clearly imaged through the GPR profiling (Fig. 6) and were confirmed by viewing through the endoscopic probe.

Figure 6.

Groundhog burrow GPR image at the Rodale field site depicting the entrance shaft, tunnel, ramp, and chamber imaged with: A) the 400 MHz antenna; B) the 900 MHz antenna, and the convergence of two parallel tunnels with the small interconnecting tunnel (highlighted with arrows illustrating the convergence) imaged with C) the 400 MHz antenna; and D) the 900 MHz antenna. The relative permittivity is estimated at 9, indicating a velocity of 0.10 m/ns.

Figure 7.

A) Map view; and B) three-dimensional view of the manual picks chosen for the identification of the groundhog burrow system through hyperbolic reflections in the 400 MHz data from Rodale. Pick choices from the north-south profiles over the active burrow are shown in black whereas pick choices from the east-west profiles over the active burrow are shown in dark gray. The light gray picks in this image represent the abandoned burrow system identified in the north-south profiles. A light gray shading is used to show the interpreted axis of the burrow. The relative permittivity is estimated at 9, indicating a velocity of 0.10 m/ns.

Figure 8.

A) Map view; and B) three-dimensional view of the manual picks chosen for the identification of the groundhog burrow system through hyperbolic reflections in the 900 MHz data from Rodale. Pick choices from the north-south profiles are shown in black whereas pick choices from the east-west profiles are shown in dark gray. The gray and light gray picks in this image represent reflections from the entrance shaft to the burrow system for the north-south and east-west profiles respectively. A light gray shading is used to show the interpreted axis of the burrow. The relative permittivity is estimated at 9, indicating a velocity of 0.10 m/ns.

Discussion

Both the 400 MHz and the 900 MHz antennas give a clear image of these key features of the burrows, with most picks coming from the depth range of 0.5 to 1.5 m (Figs. 7 and 8). However, it is clear in the three-dimensional cube image that the 900 MHz antenna can provide a more detailed depiction of the subsurface structure of the burrow, with the entrance shaft to the system as well as the two connected parallel tunnels, one leading from the entrance and the other leading to the chamber, clearly identified (Fig. 6(D)). The 900 MHz antenna survey shows the interconnection of the tunnels with greater clarity than does the 400 MHz antenna (Fig. 7 and 8). In each of the map and cube images, the scale of the burrows can be seen, with the distance between the entrance shaft to the chamber extending nearly 6 m and the side tunnels demonstrating the complexity of this mammalian burrow. The southern-most burrow imaged with the 400 MHz antenna produced reflections that were not as strong as its northern counterparts. This is likely caused by partial tunnel collapse due to abandonment.

With the small scale of the groundhog burrows, very tight line spacing is necessary to produce meaningful 3D results. The 0.1 m line spacing gives much better resolution, but is considerably more time consuming, with 10 survey lines for every meter in each direction and only 4 survey lines for every meter in each direction for the line spacing of 0.25 m. Given a survey area of 7 m by 6 m, the total number of survey lines for a line spacing of 0.1 m would be 132, whereas the total number of lines for a line spacing of 0.25 m would be 54. Although the results are much more detailed with the tighter line spacing, this alteration more than doubles the time spent in the field and significantly increases the time spent processing the data; however, this effort is well spent if the resultant high resolution subsurface imaging is desired.

Case History 4

Site Description

The final site is located within the borough of Kutztown, PA. Soil at this location is moderately sandy with small amounts of clay present. The burrow at this site was created on an eastern-facing slope and is approximately 4 cm wide at the entrance. This survey was performed in December 2015 to produce a three-dimensional image of insect burrow complexity. The burrows of cutter ants studied in Brazil were far too complex to be properly displayed in three dimensions at the GPR profile resolution obtained in 1994. As a solitary insect, the female of the burrowing bees is likely to construct a nest less complex than that of a colony of cutter ants while still maintaining aspects of insect structural complexity.

Data Collection and Processing

A GSSI SIR 3000 GPR system with 900 MHz antenna was used to survey the insect burrow location in Kutztown, PA. GPR profiles were surveyed in both the north-south and the east-west directions across the survey grid. The survey grid was 2 m east-west by 3 m north-south with a line spacing of 0.05 m and a range of 25 ns. Unidirectional surveying was used to increase location accuracy within the survey grid. Data processing in Reflexw included only distance normalization.

Results

The data were searched for hyperbolic or linear GPR reflections of increased amplitude that could correlate to the structure of the bee burrow at this site and expanded outward from the known burrow entrance. The depth of penetration is estimated at approximate 1.25 m with the 900 MHz antenna using an assumed relative permittivity of 9 based upon the soil type and matching diffraction hyperbolas within the GPR profiles to move-out calculations for different relative permittivities. Intersecting GPR profiles that display burrow reflections and the entrance shaft to the burrow system are shown in Fig. 9. The pick function of Reflexw was used to manually identify features relating to the burrow system. The survey area is shown with the picks displayed in map view (Fig. 10(A)) and in cube form (Fig. 10(B)).

Figure 9.

Bee burrow GPR image with the 900 MHz antenna showing lines: A) 0.45 m N; and B) 1.00 m E, depicting the entrance shaft and associated burrow reflections. The relative permittivity is estimated at 9, indicating a velocity of 0.10 m/ns.

Figure 10.

A) Map view; and B) three-dimensional view of the manual picks chosen for the identification of the bee burrow system through hyperbolic reflections in the 900 MHz data. Pick choices from the north-south profiles are shown in black whereas pick choices from the east-west profiles are shown in dark gray. The light gray circle represents the surface entrance shaft to the burrow system. The locations of profiles from Fig. 9 are shown highlighted as black lines at 1.00 m East and 0.45 m North. A light gray shading is used to show the interpreted axis of the burrow. The relative permittivity is estimated at 9, indicating a velocity of 0.10 m/ns.

Discussion

The entrance to the burrow system was noted in the field to be constructed down to the southwest from the surface, reaching a depth of approximately 0.5 m before diverging to a different tunnel orientation. The GPR feature related to the burrow entrance is shown in Fig. 9(B) as the truncated reflections caused by this downward sloping void at approximately 1.6N. The pick choices for this survey were selected by following the GPR anomalies from the entrance, known from the surface location, and branching outward along connecting GPR anomalies in each direction. Through this manner of mapping the tunnel, the anomaly shown on the left side of Fig. 9(B) is clearly connected to the entrance through a series of small tunnels plotted in Fig. 10. There are additional anomalies on the right side of this profile (Fig. 9(B)) which may be related to the burrow. However, to simplify the three-dimensional image, these anomalies were not given pick choices for the plot in Fig. 10. The small scale of the bee burrows required the use of a high frequency antenna as well as a very tight line spacing. The interconnection of the tunnels shown in Fig. 10 illustrates the need for such close spacing of profiles.

Discussion

Geophysical investigations of burrows have numerous applications, ranging from geohazards and groundwater flow paths to zoogeomorphology and paleontology. Geohazard studies in the realm of levee assessment indicate that half of all failures are caused by a phenomenon known as piping (Richards and Reddy, 2010). Preferential pathways of water must be detected to predict and avert failure. Examples of levee failures caused by burrowing animals have been documented (Orlandini et al., 2015; Taccari, 2015; D’Alpaos et al., 2014) and researchers have begun to respond to this with more advanced detection techniques (Borgatti et al., 2017). Zoogeomorphological investigations and paleontological research illustrate that the complexity of a burrow or burrow system is specific to the burrowing animal (Hembree et al., 2017; Storm et al., 2010; Martin et al., 2008; Hembree and Hasiotis, 2008; Smith and Hasiotis, 2008). These complexities and variations can be crucial to garnering a full understanding of the system.

This work clearly shows GPR to be useful for the purpose of mapping burrows in a variety of locations with differing soil types. Each location, however, had only a small amount of clay in the subsurface. Excessive quantities of clay can distort the GPR response due to the nature of the EM wave sent from the antenna. A high conductivity or cation exchange capacity in the subsurface, as found in most clay-rich environments, will decrease the depth of penetration of the EM pulse as shown in Equation 1 and will result in highly attenuated GPR results.

The size and complexity of the target must be considered in geophysical investigations. Cutter ants and groundhogs produce burrows of roughly the same tunnel diameter, with 5–15 cm for cutter ants and approximately 20 cm for groundhogs. Cutter ant burrow systems appear to be much more complex with networks extending far beyond that of groundhogs. Burrowing bees, however, produce slightly smaller tunnels of approximately 4 cm, with a complexity greater than the groundhogs but less than the cutter ants. Although the active cutter ant tunnel system was mapped over a length of approximately 6 m (Fig. 3) and the burrowing bee tunnels mapped within an area of only a couple square meters (Fig. 10), the inactive cutter ant burrows investigated in this work span a distance of nearly 50 m (Fig. 2). Comparing this to a distance of 6 to 10 m spanned by the active groundhog burrows (Figs. 5 and 7), it is clear that although the burrow diameter is roughly the same size, cutter ant burrows can be developed into much more expansive systems than that of groundhog burrows. The groundhog burrows were, however, shown to be more complex than a single tunnel. With parallel tunnels intersecting small connecting tunnels near the entrance shaft to the tunnel that leads to the ramp and chamber, the complexity of mammalian burrows is greater than that of their reptilian counterparts demonstrated in previous studies (Storm et al., 2010; Martin et al., 2008). These interconnections were not identified in the GPR surveys, suggesting that they are not of significantly high resolution to detect this complexity.

The depth of each tunnel system also varied between the animals investigated in this work. Although the active cutter ant tunnel system had only reached a depth of 3 m at the time of the field work at the Brazil nut plantation (Fig. 3), cutter ant burrows were demonstrated to reach a depth of approximately 6 m in their abandoned tunnel systems (Fig. 2). The groundhog burrows extend to a depth of 1.5 m whereas the burrowing bee tunnel systems extend to a depth of 1.0 m. Recall, that the depth estimates from the GPR profiles were made using Equation 2 with the relative permittivity values chosen based on known soil composition at each site combined with matching diffraction hyperbolas within the GPR profiles to move-out calculations for different relative permittivities.

Case histories 1 and 3 demonstrate an active and an abandoned tunnel system for cutter ants and groundhogs, respectively. The abandoned tunnel systems for cutter ants are highly complex and expansive (Fig. 2), whereas the abandoned tunnel systems for groundhogs are very similar in nature to the active tunnel systems. This indicates the expanding nature of cutter ant burrows, with continual excavation occurring over the course of the life of the colony.

The GPR response for the active groundhog burrow is much less chaotic than the response for the active cutter ant burrow system and for the burrowing bee. This is due to the fact that the insect systems are slightly smaller and much more complex. The abandoned burrow system for both the cutter ants and the groundhogs can be seen clearly as a hyperbolic reflection in the profile which crosses perpendicular to the axis of the tunnel system (Figs. 2 and 6(C)). However, several of the interconnecting tunnels have collapsed as the GPR response over the abandoned systems is much less distinct than that over the active systems. Since both active and inactive burrows can present hazardous conditions relative to soil competency and variable permeability and porosity affecting subsurface processes, it can be vital to identify both categories of burrows in field surveys.

The frequency of GPR antenna chosen for investigation depends upon the soil composition as well as the target dimensions and depth. In Case History 1, cutter ant burrows were investigated with 100 MHz and 500 MHz GPR antennas. Although the 100 MHz antenna provided a greater depth of penetration, the resolution given by the 500 MHz antenna was more desirable for producing profile images that could be used to identify tunnel features in the active tunnel system (Fig. 3). In the case of the inactive burrow system where only the 100 MHz antenna was used to image the subsurface, a more detailed view of the tunnel system was obtained by clipping the range of the profile and expanding the horizontal scale. Case History 3 shows the difference in GPR response between 400 MHz and 900 MHz antennas. The profile images of the main entrance shaft, tunnel, ramp, and chamber are comparable for each of these antenna frequencies (Fig. 6). However, a small tunnel that connects to a tunnel parallel to the main tunnel system is observed north of the original survey. This was identified in the more detailed 900 MHz antenna results (Fig. 8).

Antenna frequency relative to the size of the burrow or burrow system as well as the orientation of the burrow(s) must be taken into account when determining the expected GPR results from the target of the investigation. In this work, the complex cutter ant burrows were surveyed with a 100 MHz antenna at the Aruanã Site. The resulting GPR profiles from Case History 1 contain numerous hyperbolic reflections from individual tunnels within the burrow system (Fig. 2). At the Saúva Site, the reflections are much less distinct in both the 100 MHz and 500 MHz antenna results (Fig. 3), likely due to the active nature of the colony at this location. As the size of the burrows of groundhogs and cutter ants are comparable but less complex, it could be expected that the use of a 400 MHz or 500 MHz antenna used in a groundhog burrow investigation would produce a similar, higher resolution hyperbolic reflection to that observed at the Aruanã Site for a survey that crosses perpendicular to the long axis of the burrows. This is an accurate expectation at the site of Case History 3 (Fig. 6). Likewise, the bee burrows surveyed with a 900 MHz antenna are represented as distinct hyperbolic reflections when crossed perpendicular to the long axis of the tunnel and as flat high amplitude reflections when crossed parallel to the long axis (Fig. 9). However, at the site of Case History 2, the owners of the arena had filled many of the burrows with gravel to minimize potential danger to their equines. This modification creates a distortion of the GPR results (Fig. 4).

Line spacing is yet another variable that can significantly impact the success of the survey. Case histories 2, 3, and 4 provide examples of the results of four different line spacing choices, illustrating the degree of resolution that can be obtained. Although it is helpful to have high resolution data, the collection and processing of such data is time consuming. The maps and cubes of each case history (Figs. 5, 7, 8, and 10) clearly delineate the original picks from each profile, giving an example that can be culled to determine the degree of resolution necessary for future studies. While full-resolution three dimensional GPR has been demonstrated to provide numerous benefits, including the potential to improve the imaging of structures that do not run perpendicular to the GPR survey lines (Dogan et al., 2011; Grasmueck et al., 2005), the results of this work are presented with two-dimensional treatment to better illustrate the potential resolution of the different line spacings. Likewise, the perpendicular survey lines for these case histories provides an example of the directional dependent results. Whereas the survey lines that cross perpendicular to the long axis of a tunnel may obtain hyperbolic reflections of notable amplitude, the survey lines that cross parallel to the long axis of a tunnel may not provide a distinct enough anomalous GPR signal to warrant note. This is demonstrated in Case History 3 (Fig. 7) as the abandoned burrow did not produce notable GPR anomalies in the east-west survey lines (parallel to the long axis) but was visible through the GPR survey lines in the north-south direction. For burrows of unknown orientation, this highlights the importance of surveying more than a single direction.

In Case History 2, the goal of the survey was to identify any anomalies indicative of a groundhog burrow and the line spacing chosen for this was 0.6 m. The distribution and density of anomalous responses are shown in the pick file for this line spacing (Fig. 5). Comparing this resolution to that obtained with a line spacing of 0.25 m using the 400 MHz antenna (Fig. 7) and 0.1 m using the 900 MHz antenna (Fig. 8) it is clear that the tighter line spacing in Case History 3 provides a much more distinct image of the subsurface structure of the burrow system. This resolution is gained at a significant loss of time both in the field collecting the data and at the computer processing the data. The tighter line spacing was necessary in the case of Case History 3 as a three-dimensional depiction of the burrow complexity was the goal of the investigation. In Case History 2, however, the larger line spacing did not detract from the goal of the survey, which was to locate anomalies indicative of groundhog burrows. Case History 4 had an even tighter line spacing (0.05 m) and a high frequency antenna (900 MHz). Due to the small nature of the burrowing bee tunnel diameter, determining an accurate connection of the GPR anomalies between lines would have been impossible without this high-resolution imaging. The results show the importance of line spacing in three-dimensional GPR imaging of subsurface structures. A line spacing of 0.25 m was used for the 400 MHz antenna, whereas a 0.10 m and a 0.05 m line spacing were used for the 900 MHz antenna.

The presence of noise in the data is something that must be considered when choosing line spacing. Figure 8 offers a clear image of the location of the active groundhog burrow, but there are several profiles which contained GPR reflections whose source might be interpreted erroneously as a tunnel if not properly analyzed with adjacent and perpendicular line confirmation. This is demonstrated by Fig. 8 in the scattering of picks near but not connected to the main burrow tunnel and Fig. 9(B) in the anomalies on the right side of the profile. GPR anomalies that are discontinuous across lines may be shown as unrelated to the burrow (Fig. 8) or potentially tracing out a more complex burrow system (Fig. 9(B)). Although anomalies may match the size and shape of the expected burrow reflection, discontinuity with the rest of the reflectors identified in the three-dimensional image can show them to be unrelated to the main burrow. When choosing line spacing, it is vital to choose a spacing that will allow interpretations to be made that distinguish between noise and target.

Conclusions

GPR is demonstrated as a useful tool for mapping animal burrows in three locations of differing soil type. Low frequency antennas (100 MHz) can produce images of the subsurface burrows or burrow systems of cutter ants and groundhogs, whereas higher frequency antennas (400 MHz, 500 MHz, and 900 MHz) provide higher resolution within the range of these burrows or burrow systems. Three-dimensional surveying can be used to produce high resolution images of subsurface features. The appropriate choice of line spacing is vital to the accurate identification of subsurface burrows or burrow systems. A line spacing of 0.6 m can produce an image of the distribution of the burrow network, whereas a tighter line spacing (0.25 m for 400 MHz or 0.10 m for 900 MHz) can aid in mapping the small connections between adjacent tunnels of groundhog burrows or (0.05 m for 900 MHz) burrowing bees. Surveys of this nature can have applications in hazard assessment, agriculture, fluid migration, contaminant fate and transport, zoogeomorphology and paleontology.