Non‐invasive surveys to assess animal burrows as potential causes of levee instability

Animal activities threaten levee stability, especially during flood events, increasing levee failure risk. Indeed, while dens alter hydraulically‐induced failure mechanisms, it is difficult to locate them and to know in advance if critical conditions may be reached. This paper demonstrates that it is possible to perform nondestructive surveys to assess the risk of burrow animal activities in levee systems using a ground penetrating radar (GPR) for effective levee structure control, being GPR a nondestructive, expeditious, and cost‐effective geophysical technique. Here, a flood‐prone area in Sicily, where traditional hydraulic phenomena did not justify past levee failure events, is considered a case study. Through a visual inspection survey, several animal holes were recovered. GPR measurements allowed us to explore the levee interior and to detect the presence, distribution, and configuration of dens inside them. It was found that a complex burrow system crosses the entire embankment. Porcupine was recognized as a builder of such cavities. Although, previous studies established the porcupine presence in semi‐humid environments, for the first time, here we provide evidence of the development, articulation, and extension of porcupine burrows inside the levee system.

development of different animal species, especially burrowing animals.Past studies indicate the connection between levee failure mechanisms and animal activities (Bayoumi & Meguid, 2011;Camici et al., 2010;FEMA, 2005;Orlandini et al., 2015;Palladino et al., 2020;Taccari & van der Meij, 2016a;Taccari & van der Meij, 2016b).Nevertheless, these studies do not give sufficient indications about quantifying such an impact.In addition, after studying the past events it is not easy to reconstruct the actual development of the levee breach because the evidence of its origin is generally covered up during failure (Hagerty, 1991).
Many animal species construct complex dens and articulated tunnel systems (Reichman & Smith, 1990), such as badgers, porcupines, foxes, coypus (nutrias), and crayfish.Animal tunnels may have different shapes and dimensions depending on which animal has dug and for which reason.The Federal Emergency Management Agency reported 23 main species among those posing a threat to earthen dams (FEMA, 2005).Many of these species share common characteristics, but the pattern and size of burrows and the damage severity they cause to earthen structures could substantially vary (Bayoumi & Meguid, 2011).Thus, there is a significant lack of data to fully comprehend levee failure related to biota activities and, in the past years, the problem was unfortunately underestimated.
Levee monitoring involves tracing any sign of potential weakness by visual inspections and requires unsafe destructive penetrations.This conventional geotechnical method is not a suitable technique, either in terms of time or cost efficiency (Di et al., 2010).Additionally, it provides evidence of levee conditions only in a limited area.Nowadays, several stakeholders believe that determining the levee health status in a speedy and nonintrusive way over a large area is of primary importance for the efficiency of our flood defense system (Di et al., 2010;Loperte et al., 2011).Ground penetrating radar (GPR) is a promising geophysical technique in terms of survey velocity, structure nondestructiveness, and cost-effectiveness to perform quick and easy inspections of large portions of levee body (Di et al., 2010;Szynkiewicz, 2000).Since the early decades of the twentieth century, GPR has been used to detect buried objects, and since 1926 the GPR application has been used to determine the depth of rock strata using the time-of-flight method (Manacorda et al., 2015).Even if engineers use the GPR as a nondestructive tool for mapping underground characteristics (Di et al., 2010), GPR applicability to river levee analysis is not supported by extensive literature.Only since 2000, Szynkiewicz has appraised the GPR helpfulness for monitoring and evaluating earth levee integrity.Biavati et al. (2008) pioneered the GPR capability in different sections of a levee, their preliminary results proved the appropriateness of this technology with different soil types.Moreover, Chlaib et al. (2014) used GPR to assess the levee structure integrity, and they discovered its usefulness to determine subsurface animal burrow positions.Other manuscripts used GPR to locate animal burrows, such as Cortez et al., 2013, Kinlaw et al., 2007, and Nichol et al., 2003; in particular, such authors investigated burrows of the pocket gopher, gopher tortoise, and badger respectively.However, to the author's knowledge, no previous study has applied GPR to characterize the geometry and the dimensions of animal burrows, with special regard to porcupine dens.
In the present work, we explore the capability of a nondestructive methodology for assessing flood risk related to animal digging activity in a levee system.In particular, a contribution to understanding failure mechanisms due to biota activities and assessing GPR applicability to recover the geometry and structures of the burrows is offered.To this aim, analytical and numerical models are coupled with field surveys and GPR measurements.
The results of the investigation provided evidence of large mammal activities in the levee system with specific reference to a case study in Sicily, which is a region characterized by a semi-arid climate.Starting from the reconstruction of a recent flood event that was triggered by non-obvious causes, through visual examination in situ, we discovered the presence of porcupine holes in the levees.Till today, this species has never been considered detrimental to levee stability.
To evaluate the distribution of faunal burrows and their configuration, control of levee structures and nondestructive investigations were performed using GPR.The potential of this technology was tested to identify the species and their burrowing behavior within Sicilian levees.
In the next section, the applied methodology is summarized.In the third section, geomorphological characteristics of the analyzed area and a reconstruction of the 2012 flood event are described.The fourth section focuses on the analyses of the results, concerning an assessment of the levee failure mechanisms occurring during that event, the evidence of burrowing animal activities, and the GPR survey investigation.Finally, the main results of the work are summarized, and final considerations are given.

| Hydraulic assessment of levee vulnerability
After a flood event, one of the main issues from the technical and scientific points of view is to identify the critical condition that triggered failures.This paper focuses on the most common hydrological and geotechnical failure mechanisms, that is, overtopping and seepage, which could have occurred during a specific flood event.
This paper presents numerical simulations created by HEC-RAS software (Brunner et al., 2020;Gary, 1995), which estimate the water profile to verify levee overtopping.To this aim, the river planimetric development and its control cross-sections have to be modeled, including structures that cross the river, such as bridges.The computational grid is enhanced to improve simulation accuracy by linear interpolation between every two consecutive cross-sections.For the sake of safety, steady flow profiles at the peak flow rate were computed, by considering mixed flow regimes, that is, both superficial and subcritical flow.The Soil Conservation Service Curve Number method (SCS-CN) (Mishra & Singh, 2003) and the rational method are applied and compared to evaluate the flood hydrograph and to determine the peak flow rate.While all simulations could be affected by a margin of error related to limited knowledge of realistic conditions during the flood event, here, the primary source of error is likely associated to the uncertain value of hydraulic roughness.As a protective measure, the lowest Manning coefficient related to the channel river condition is assumed.
Concerning the seepage phenomenon, it may induce the transport of soil particles of the embankment, or its foundations, altering the levee stability; thus, levee failure may be due to different types of internal erosion, generally through four possible mechanisms: concentrated leak, backward erosion, contact erosion, and suffusion (ICOLD, 2017).Erosion progression mainly depends on the hydraulic and mechanical conditions (ICOLD, 2017).Unless forces are reduced, the initiated erosion continues until the formation of pipes and breaches in the levee body.
Although the problem of piping failure mechanism is extremely interesting, this paper focuses only on the overall seepage processes through the embankment by looking for the critical condition that triggers levee piping, leading to extreme erosion resulting in the inevitable levee collapse (Camici et al., 2017).This critical condition is reached when the phreatic line, associated with the temporal evolution of the seepage process, achieves the landside surface.Until the phreatic line is contained inside the levee body, the hydraulic failure of the levee will not occur.Otherwise, a threatened status will occur.Michelazzo et al. (2016) suggest simultaneously applying Marchi's (1961) and modified Green-Ampt's (Pistocchi et al., 2004) models to obtain a conservative evaluation of the levee vulnerability to seepage.These models start from quite different assumptions which complement each other.In particular, Marchi's model is the most realistic approach for describing the seepage phenomenon in the lower part of the embankment; contrariwise, the modified Green-Ampt's model is the most realistic one in the upper part of the levee, specifically near the crest (Michelazzo et al., 2016).Moreover, this second method is unrelated to the depth of the undisturbed water table.

| Visual investigation and GPR survey
Wildlife interacts with earthen levees as if these were natural fields or forests.Thus, a significant step toward fortifying a levee against the effect of nuisance wildlife damage is observing clues left by animals (FEMA, 2005).
First, as suggested by FEMA (2005), the inspections are performed by walking back and forth across the slope, utilizing zig-zag patterns, to check the entire levee surface and individuate evidence of levee instability.
From a simple visual levee inspection, it is impossible to measure the animal burrow geometries inside the levee, so it is problematic to evaluate the risk associated with their presence.To study animal burrow architectures and extents, Roa et al. (2014) injected cementbentonite mix inside holes and then excavated the levee, bringing to light the entire system.Although they managed to document the volume and geometry of the burrows, in the end, the levees were destroyed.Thus, their methodology is too invasive to be re-proposed here to evaluate levee health.Here, GPR scans of the embankment at the levee crests are carried out for a rapid and non-destructive survey of the levee health state.The advantages of this method relate to the opportunity to operate over an extensive zone and to check the burrow evolution periodically.Since GPR works using propagating and reflecting electromagnetic waves inside the ground, the data acquired are a function of the round-trip transit time of the pulse and its reflection (Manacorda et al., 2015).The survey results are more reliable the smoother the topographical coverage and the drier the penetrated material are (Reynolds, 1997).

| GPR information and calibration
The GPR employed in the present investigation is an IDS GeoRadar (RIS MF Hi-Mod model), multifrequency, multichannel array system.This GPR captures subsoil three-dimensional images, improving hole detection.It has two antennas in one box, one of 200 MHz and the other of 600 MHz, which allow real-time visualization of the ground structure at two different resolutions.The GPR provides a real-time reconstruction of the soil sections of deep and shallow antennas on the same screen.The image analyses allow detection trends of discontinuity and inhomogeneity characterizing the investigated site.In this work, the penetration depth reaches 2 and 4 m using 600 and 200 MHz antennas, respectively.Generally, the penetration depth may vary depending on soil conditions.
An instrumental configuration set-up is needed to start the survey to calibrate the system on the investigated soil.This operation guarantees the required depth survey and optimizes signal quality.
The GPR automatically generates a cartographic map containing cross-section and planar view results.Data processing and semiautomatic layering reconstruction are performed using the GRED-HD software, provided by IDS Georadar.The results can be exported in GIS and CAD formats.

| Characteristics of the levee defense system
The Dirillo River basin is located in the SW of the island of Sicily and outflows into the Mediterranean Sea, near the city of Gela.It flows from the NE to the SW, from 986 m.a.s.l. to zero m.a.s.l.Its channel divides the border between the Caltanissetta and Ragusa districts.The river length is 54 km with a 740 km 2 catchment area.
The studied area is located downstream of the SS115 road bridge.This is the only part of the Dirillo River protected by levees.Thus, the SS115 road bridge is the upstream limit of the catchment area, and the input flood hydrograph is estimated in this section.In the upstream part of the catchment area, there is a Ragoleto dam.Its presence represents a hydraulic discontinuity of the whole basin.Figure 1 shows the catchment area at the SS115 road section (A) and the one bounded by the Ragoleto dam.In the following analyses, we focus only on the characteristics of the downstream basin (A).Nevertheless, the Ragoleto dam hydrological effects are considered through the reconstruction of its outflow discharges during the flood event.
The catchment area characteristics are the following: the surface is 230.65 km 2 , the perimeter is 138.38 km, the average slope is 17.51%, the maximum elevation is 903.82 m a.s.l., the minimum elevation is 24.95 m a.s.l., and the average elevation is 505.28 m a.s.l.The morphometric characteristics of the main channel related to the outlet are the following: the river length is 50 km, the average slope is 0.96%, the maximum elevation is 503.82 m a.s.l. and the minimum elevation is 24.95 m a.s.l.
All the areas near the river are characterized by intensive agriculture.Thus, the cultivation fields achieve the levee toe, increasing the flood risk for the zone.In the upstream area, the valley slopes are steep, and they are made up of calcarenites and sands.The lower part is characterized by clay-sandy soils.Currently, erosive processes are concentrated on the slopes further upstream, while alluvial material settles in the valley areas and within the river channel.Nevertheless, the territorial structure has a generally stable morphology, with limited and localized evolutionary phenomena.A delta environment with an estuary characterizes the outlet to the sea.A coastline regression is undergoing due to the reduction of incoming sediment (PAI, 2004).
Downstream of the SS115 road, the floodplain is protected by earth levees.The area is intensively farmed, mainly through greenhouses.In the 1930s, a levee system made up of local materials consisting of sand, silt, and clay was built to prevent flooding and overflow (Figure 2).In 1948In , 1949In , and 1951 As highlighted by historical analyses, before 2012, the last repair works were performed in 2008.These works were done to remove deposited sediments from the main channel but not to enlarge the levee sections, and the surveyed levee transects are the same as those of the 1930s original project, as shown in Figure 2. The data and cartographic material available from the 2008 repair work of the control structures are the most complete and upto-date information about the configuration of the Dirillo River before the 2012 flood event.However, the evaluation of levee height variation is complex because, over the years, they may have suffered subsidence.Additionally, the animal burrow presence likely deteriorated these flood protection structures, therefore endangering their resistance.

| The 2012 flood event
The Dirillo River basin is frequently affected by floods, and the 2012 event is the emblem of the flood events occurring in this area.It caused tremendous physical damage with consequential economic and social disruptions.The stream broke the river levees at different points.Through landowners' historical memory and reports drafted by the Civil Protection Department, locations, where levee overflows occurred, were determined, as shown in Figure 3.
First, a reconstruction of the flow rate during the event is given.Next, the vulnerability to overtopping and seepage phenomena is examined in the case of the undisturbed levee.
Concerning rainfall data, records of the March 10 and 11, 2012 are considered, measured at the rain gauge stations of the Water Observatory of the Sicily Region in Acate (62 m a.s.l.), Mazzarrone (303 m a.s.l.), and the Ragoleto dam (331 m a.s.l.) (see Figure 3).
Table 1 reports the rain event characteristics recorded at the three rain gauges and gives the maximum rainfall depth for different time durations.
After determining the influence areas of all the stations, and computing the weighted average of the collected rainfall data, the rainfall causing the flood event is obtained.The influence area of the rain gauge stations and their influence percentage over the total basin are estimated: the Acate influence area is 46.22 km 2 , which corresponds to 20.05% of the catchment area; the Mazzarrone influence area is 71.16 km 2 , which corresponds to 30.85% of the catchment area; the Ragoleto dam influence area is 113.26 km 2 , which corresponds to 49.10% of the catchment area.The comparison between the SCS-CN method and the rational method, to evaluate the flood hydrograph, is not reported here for the sake of brevity.However, Figure 4 shows the final flood hydrographs.In particular, the flood hydrograph at the SS115 road outlet is adjusted since the hydrographs obtained by both the SCS and the rational method, considering their sub-basin catchment area, must be increased by the dam flow released, as recorded through reservoir water levels during the 2012 flood event.Moreover, the dam outflow is offset in time, considering the lag time needed to reach the outlet.

Rain Gauge station
Hydrographs shown in Figure 4 are the followings: the runoff hydrograph estimated with the SCS method (SCS) of the contributing sub-basin without dam outflow; the runoff hydrograph evaluated with the Isochrone method (I) for the same sub-basin; the outflow hydrograph of Ragoleto dam estimated through water level measurements at the reservoir (DO); the flood hydrograph evaluated as the sum of the runoff hydrograph obtained by the SCS method and the lagged dam outflow hydrograph (SCS + DO); the flood event hydrograph estimated as the sum of the runoff hydrograph obtained by the Isochrone Method and the lagged dam outflow hydrograph (I + DO).
The peak runoff rate Q p , calculated as the sum of the runoff hydrograph and the dam outflow hydrograph, is about 262 m 3 /s for the SCS method and about 236 m 3 /s for the Isochrone method.
The SCS method provides the highest value of the peak discharge.Therefore, the combined hydrograph between the SCS method and dam outflow gives more conservative estimates for the following analyses.

| Assessment of possible hydraulic traditional failure mechanisms
To evaluate the overtopping phenomenon as a cause that triggered the failure mechanisms during the 2012 flood event, steady-state HEC-RAS numerical simulations are performed.
River section geometry was recovered by field topographic measurements carried out in 2008 by the local Office of the Water Authority, which realized maintenance works of the Dirillo river and levees.Figure 5 shows the planimetric view of the river and the location of the control cross-sections.The maximum distance between interpolated cross-sections is 1 m.The river slope is about 3‰. Figure 5 also represents the location of different structures that crossed the river, such as bridges, corresponding to sections 37, 29, 22, 18, and 12.The geometrical characteristics of such cross structures were measured during a field survey.
The levees were repaired in 2008, so it is safe to assume that in 2012 vegetation had already started to grow.Since the literature suggests a Manning's coefficient between 0.10 and 0.03 s/m 1/3 for natural vegetated channels (Arcement & Schneider, 1984), a coefficient equal to 0.03 s/m 1/3 has been considered for safety sake.
Regarding the steady flow analysis, the flow rate considered is equal to the peak runoff rate Q p evaluated with SCS methodology at the outlet, about 262 m 3 /s.Noting the slope of the main channel, its length, and the flow rate values, we feel that it is appropriate to regard the upstream boundary flow condition as uniform flow.
Figure 6 shows the calculated water profile.River levees seem to not be always adequate to contain the water flow.In particular, the bridge structures reduce the river section, and the elevation of the levees is not high enough in the correspondence of the Feudo Arancio Bridge and the diversion dam.Consequently, close to these two structures, the water overflows the river banks.Additionally, Figure 6 shows the locations pinpointed from the landowners' memory of the 2012 flood and the reports drafted by the Civil Protection Department as points with levee overflow and flooding of the flood plain (see also Figure 3).
Results highlight some likely overflow near the crossing structures, specifically Feudo Arancio Bridge, Railway Bridge, and Division Dam.In the HEC-RAS simulations, these structures are indicated as sections C, D, and E, respectively.Nevertheless, during the 2012 flood event, other overflows are recorded at unexpected places, such as points A, B, and F in Figure 6, corresponding to sections 33, 32, 24, and 14, in Figure 5, respectively.
In particular, the freeboard, defined as the distance between the levee crest and the free surface, is about 1.42 m on the left side and 2.71 m on the right side at section no.33, about 0.42 m at section no.32, about 0.59 m at section no.24 and about 1 m at both sides of section no.14.
As can be noticed, the sections where overtopping occurred are all near the bridges (C, D, and E sections).Instead, the water profile should not have overtopped the levees at other points along the river (sections A, B, and F).These discrepancies between the actual event of 2012 and the HEC-RAS simulated event could be due to other factors.Therefore, overtopping might not be the only mechanism triggering levee failures during the 2012 flood event.
Afterwards, all Dirillo cross-sections have also been studied to evaluate the seepage mechanisms that could have triggered levee failures.
Considering the water level duration in the Dirillo River, the variations of the phreatic line inside the levees are calculated.The flood event started at 7:00 on the March 10, 2012 (the first hour in Figure 7) and finished at noon on the March 13, 2012.
Regarding the soil characteristics of the levee system, it is composed of sand and limestone, and it has low infiltration rates.Therefore, it is safe to assume a porosity equal to 0.4 and a hydraulic conductivity equal to 10 À5 m/s.The Strickler coefficient used is 30 m 1/3 s À1 , being a conservative estimate.
The analyses show that the phreatic line may never have reached the landside slope of the levee crosssections.To provide an example, Figure 7 reports the results of both Marchi's and modified Green-Ampt's models obtained for section E (see Figure 6), showing the evolution of the phreatic line during the flood.As it can be seen, such a line never crosses the land slope of the levee.Thus, the seepage phenomenon seems not to be a plausible mechanism to explain the levee failures during the 2012 flood event.
It is worthwhile to mention that the hydrological and hydraulic modeling outputs are affected by uncertainty, which may result from different sources of errors (e.g., input data, model structure, and model parameters).
Despite that, results indicate that the main failure mechanisms during the 2012 flooding event in the Dirillo River floodplain at some sections cannot be explained by overtopping or seepage.However, since the investigated area of Dirillo River is likely to be a burrowing animal site, it stands to reason that some levees were already deteriorated by animal activities before the flood event.This condition may have triggered local collapses.

| Visual inspection results
The goal of the present inspection is to check the entire levee surface.The Dirillo levees are inspected between the SS115 road and the SP31 road.Specifically, as shown in Figure 8, from SS155 road to 2 km downstream (Feudo Arancio Bridge), the right levee of the river is observed; from that point to SP31 road, the left levee of the river is monitored.
Unfortunately, the presence of dense vegetation and the steep inclination of the levee slope does not allow investigation of the whole riverside slope.However, eight suspect holes were found and inspected.Figure 8 shows the dens and their position.
Hole number 1 is located on the right levee of Dirillo River, lightly hidden under the vegetation.The trail of the "homing" (e.g., well-shaped den entrance), the recast rubble material outside of the den, and the rest of the animal activity (e.g., food, feces, etc.) demonstrate that it is used.Dimensions and entrance conformations of hole number 2, located on the riverside of the same earth levee, suggest that the same animal species lives in these complex burrow systems.The holes from number 3 to number 8 are easily identifiable on the landside slope of the left levee.That is possible due to the maintenance work of landowners.The levee landside is located on Feudo Arancio's property, and the owners have committed themselves to keep the crest and the slope of the levee always tidy and clear.
The holes are located on the middle-upper part of the levee slope.They are all near the levee crest and never below 2.00 m from it.All this suggests that the burrowers should be terrestrial animals.Terrestrial animals prefer to inhabit or hunt in the levee crest area.In this way, the entrances of their burrow system are not below the water level.On the contrary, amphibian animals, such as nutria, prefer the lower slope area.
As is the case of holes number 1 and 2, these dens are actively used by the hosts.
Moreover, even though the den entrances have different shapes, the tunnel dimensions are almost always the same: the diameter is about 0.30 m and their depth is longer than 2 m.As shown in Figure 9, during the in situ investigation, a 2 m-long wooden ruler is completely inserted into the holes without encountering any obstacles.
The in situ investigation suggests that all the burrows belong to the same animal species.The hole dimensions and their positions on the slope levee suggest that the burrowing animal is a medium-sized mammal.The quills and pawprints found at the entrance (see Figure 9) suggest that individuals of Hystrix Cristata (porcupine) are responsible for the presence of burrows within the investigated levees.The porcupine is a medium-sized animal, and it is the largest species among Italian rodents, as reported by Pigozzi (1986), who defined average weight and length: male porcupines weigh about 10.1 ± 1.4 kg and have a total length, head-body length, of about 70.8 ± 2.6 cm; female specimens weigh about 11.4 ± 1.7 kg and have a total length of 73.3 ± 3.3 cm.
F I G U R E 8 Dirillo levees inspected: Levee on the right of the river from SS115 road to Feudo Arancio Bridge; left levee of the river from Feudo Arancio Bridge to SP31 road; location of the eight discovered holes.
F I G U R E 9 Field evidence of the presence of burrowing animals: porcupine footprint and quill near hole no.6; and burrow dimension analyses.
Porcupines are terrestrial mammals covered in long spines or quills, their distinguishing feature, who live in family groups in their complex burrow systems (Bayoumi & Meguid, 2011).These terrestrial mammals are herbivores, feeding on leaves, herbs, twigs, roots, bark, berries, fruit, and even farmers' crops.They are not solitary, but they usually live in small family groups of an adult pair and their offspring (Felicioli et al., 1997).This family group digs its complex tunnel system where they spend the daytime.Generally, the burrows are hidden in areas characterized by dense vegetation (Mori & Assandri, 2019), and their dens can be up to 15 m long with a 2 m-deep living chamber and one or more openings (Mendelssohn & Yom-Tov, 1999;Pigozzi, 1986).The female will move to a separate part of the tunnel system when giving birth to new young, building her nesting chamber, where she stays alone (Corsini et al., 1995).However, it is also common to see porcupines alone.Porcupines do forage for food alone, usually at night-time when they are most active; they will then return to the family den for the day.These mammals can walk long distances each night in search of food (Sever & Mendelssohn, 1991).Sometimes, they can travel up to 2.5 km per night (Pigozzi, 1986;Sever & Mendelssohn, 1991).
Consequently, although the habits of this solitary species may, in principle, be judged as not compatible with the intense agricultural use of the site, the porcupine's physical features and their behavioral characteristics could justify the presence of tunnels dug by them within the Dirillo River levee.During the daytime, they could hide in the riverside slope reeds, and during the night, they may go out undisturbed to look for food on the landside.
Having identified the existence of animal dens by visual inspection, the position and paths of the burrow tunnels inside the levee are then reconstructed by employing a GPR survey.

| GPR analysis
The identification of numerous holes suggests the presence of a complex burrow system developed inside the Dirillo River earth levees.These tunnels may have also affected the earth levee during the 2012 flood event.
The GPR is calibrated on a planar surface, and the soil's electromagnetic characteristics are verified.Moreover, the calibration test allows for obtaining information on wave propagation speed and permittivity.The reflection time is converted into depth by knowing the propagation speed.A time window, which represents the penetration depth, is selected.The chosen value is a function of the antenna's central frequency and soil conditions.Keeping in mind that the purpose of the investigation is to identify burrows up to a depth of about 4.00 m (the height of the levee), GPR is set to a 150 ns time window, which allows for a 6.00 m deep survey is considered.
The investigated levee sections are chosen considering the presence of nearby holes.The studied levee sections are collocated in the correspondence of two couples of holes: 1-2 and 5-6 (see Figure 8).
The inspected levee section is 3.50 m wide (Figure 10a).The survey area is 3.00 m long over the levee crest (Figure 10a).A 1.00 Â 1.00 m grid is considered at the crest of the structure to perform the survey (Figure 10b).Through the GPR surveys, 16 profiles along the streamwise and cross-section direction are acquired for each section.The survey procedure performed on-site is schematized in Figure 10, which shows the main phases of the GPR survey and the final reconstruction of the animal's den path inside the levee.When a cavity is detected, its position is marked with red spray paint on the ground (Figure 10c).Identifying a sufficient number of discontinuity points allows for reconstructing the internal levee burrow path (Figure 10d).
The GPR survey allows the evaluation of the tunnel paths, depth, and dimension inside the levee.
The burrows recovered are located about 1 m under the crest surface.The tunnel dimensions are variable, but they rarely have a diameter smaller than 0.35 m.The GPR survey confirms that porcupines populate these dens, considering the tunnel dimensions and patterns inside the levee.Table 2 reports measured burrow's geometrical characteristics.
The tunnels investigated by the GPR survey correspond to holes number 1, 2, 5, and 6.During the visual inspection, we do not find the riverside entrance (hole 2) because the river vegetation (mainly reeds) makes the riverside slope inaccessible.The GPR survey allows the discovery of the connection between hole number 1 to hole 2.
Figure 11 shows the surveyed levee structure as obtained by the GPR in correspondence to burrow 1.The cross-section GPR scan shows the clear presence of hyperbolas, emphasized in Figure 11 by the arrows, which indicate the presence of two hairpin bends to reach the den.The tunnel that connects both holes is shaped like a wave, and it is located about 50 cm below the levee's crest level.Moreover, its path inside the levee points forward in the river flow direction.Figure 12 proposes a 3D reconstruction of the burrow inside the levee.
Figure 13 displays a reconstruction of the burrows inside the Dirillo River levee for holes 5 and 6.These holes are located on the landside slope of the embankment and are connected by tunnels to each other.The tunnel system connects these two holes with a third entrance positioned on the riverside of the levee slope.These dens are located on the higher part of the levee, approximately 50 cm below the crest level.In this case, the den tunnel system has one main entrance on the riverside slope with a single straight tunnel about 1 meter long, which splits into two paths, making a "y" shape, and finishes with a double exit on the landside slope.The slope of the dens inside the levee is oriented opposite to the river flow direction.The present research originates from the need to understand the actual risk of the earthen levee stability affected by burrowing animal activities.The levee failure mechanisms in the presence of biota activities are complex, and the assessment of their impact on levee stability is unclear.GPR is tested as a quick and effective way to detect the presence of biota activity and to manage levee systems.
Preliminary, possible mechanisms that triggered levee failures in the Dirillo River (Sicily) during the 2012 flood event were analyzed.The levee vulnerability was evaluated concerning overtopping and seepage failure mechanisms.Such analyses show that the reconstructed 2012 flood event appears not extreme enough, in terms of both peak river stage and duration of high river stages, to trigger the overflow phenomenon at so many sections as it occurred in reality.Furthermore, the river stages are such that the seepage phenomenon is unlikely.To the authors' F I G U R E 1 2 Reconstruction of burrow path inside the levee through GPR survey.Investigated holes 1 and 2.
F I G U R E 1 3 Reconstruction of burrow path inside the levee through GPR survey.Investigated holes 5 and 6. knowledge, no prior reports of detailed animal burrows inside Dirillo River levees were available, and no link was made to levee failures during floods.
Field surveys were performed to determine additional factors that could harm levee stability.Porcupine holes were identified along the 8 km of landside levee surfaces by a preliminary visual inspection.Their diameter is about 0.30 m, and their depth is longer than 2 m.While it is known crested porcupines inhabit the whole Sicilian island, the dens of these terrestrial animals have never been reported in artificial levees before the present study.The dense vegetation of the river that makes the area access difficult and the food presence on agricultural neighboring land areas are suitable habitats for this animal species.Quills and pawprints recovered close to the den entrance confirm the presence of porcupines and the active use of burrow tunnels.
A fast, nondestructive, and optimized survey to investigate biota tunnel configuration was performed using the GPR to measure animal burrows inside the levee.The GPR analyses revealed that the Dirillo River levees are crossed by a complex tunnel system, which may affect the levee stability.The investigation results allow us to quantify the depths and paths of dens inside the levee.Moreover, they reveal that den tunnels start as a single straight tunnel and, inside the levee, they split into more than one tunnel connecting the riverside slope to the landside slope.This evidence, never revealed by previous works, highlights the potentially destructive impact of den presence inside the levees.During flood events, these tunnels may trigger preferential infiltration and channeling processes and, in the worst case, an intense erosion that could trigger the failure of the overall structure.The non-invasive investigation through GPR provided fundamental guidance for monitoring the earthen levee.
In conclusion, the analysis of the flood event highlighted the relevance of the problem related to the burrowing animal activities on the levee stability, even in Sicily.The field investigations revealed intense fauna activity.The GPR has been proven to be a suitable, expeditious, and cost-effective tool to assess the health state of the embankment sections and investigate tunnel den distributions and geometries.The methodology proposed can effortlessly be performed on long stretches of river levees.Additionally, the GPR investigation could be used to update flood risk maps and define the localized area where sustainable restoration techniques should be applied.
Periodic monitoring is necessary to safeguard the levee integrity.Several effective remediation techniques, such as Cased Secant Piles, Jet-Grouting, cement, chemical grouting, and micropiles, can be adopted to secure levees affected by burrow networks.Although these techniques provide levee stability, they are unsuitable and require relevant investments to be implemented.Composite geotextile materials could be a suitable and eco-friendly remediation technique (Pennisi, 2020).However, future works are necessary to assess the resistance of this technique to the action of rodents like porcupines.

1
Location of Rain gauges (red points) and the catchment area: (a) Catchment area from Ragoleto dam to the outlet at SS115 road; (b) Catchment area upstream of Ragoleto dam.
, the levees were damaged by several flood events and were rebuilt.The defense structures were designed considering a maximum flow rate of 200 m 3 /s.Recently, flooding events occurred in 2005, 2006, and 2012.During the 2005 flood event, the levees broke on the left side of the river, while in 2006 and 2012, levee failures occurred at different sections along the river.Due to extensive damage, in 2014, the Civil Protection Department of the Sicily Region identified Dirillo Valley as a recurring flood event area.

F
I G U R E 2 (a) Historical map of authorized hydraulic works to protect the Dirillo River valley in the 1930s and (b) section of the original project (adapted from Consorzio Idraulico Fiume Dirillo, Comune Vittori) (c) 2008 repair work project (cortesy of the Department of Genio Civile Ragusa and ICARO Ecology s.r.l. of Gela).

F
I G U R E 3 Reconstruction of the 2012 Flood of the Dirillo River and location of levee failure points-A, B, C, D, E, F (information provided by Department of Civil Protection of Sicily Region).T A B L E 1 Characteristics of the rain event of March 10 and 11, 2012 recorded by Acate, Mazzarrone, and Ragoleto dam stations and maximum rainfall value for the duration of 1, 3, 6, 12, and 24 h, (Water Observatory of Sicily Region).

F
I G U R E 4 Runoff rate hydrographs of the catchment area with SS115 road outlet: (a) SCS = the runoff hydrograph estimated with the SCS Method; (b) I = the runoff hydrograph estimated with the Isochrone Method; (c) DO = the outflow hydrograph of Ragoleto dam; (d) SCS + DO = Sum of the runoff hydrograph estimated with the SCS method and the dam outflow hydrograph; (e) I + DO = Sum of the runoff hydrograph estimated with the Isochrone method and the dam outflow hydrograph.

F
I G U R E 5 Planview of Dirillo River and location of control cross-sections obtained from the project of 2008 restored and by field survey.F I G U R E 6 Water profile in a steady flow simulation of Dirillo River, considering a flow rate of 262 m 3 /s, and the position of the overflow points during the 2012 flood event (see Figure 3 for plan view location).

F
I G U R E 7 Evolution of the phreatic line inside the levee during the flood.Each color of the phreatic line corresponds to a precise stage of the flood event.The color bar on the right indicates the time of the 2012 flood event starting from 7:00 on the March 10, to noon on the March 13.The color bar on the left shows the water level elevation during the whole flood event.

F
I G U R E 1 0 GPR survey procedure for holes 5-6: (a) GPR position on the levee crest; (b) measuring area for the GPR scans; (c) Cavity detection; (d) Reconstruction of burrow path inside the levee (red line painted on the ground).
Measured geometrical characteristics of dens inside the levee system.
T A B L E 2 F I G U R E 1 1 GPR scan: Burrow path inside the levee and GPR graphic visualization with 200 MHz antenna analysis (burrow 1).