Characterisation of Fractures and Fracture Zones in a Carbonate Aquifer Using Electrical Resistivity Tomography and Pricking Probe Methodes

Position, width and fragmentation level of fracture zones and position, significance and characteristic distance of fractures were aimed to determine in a carbonate aquifer. These are fundamental parameters, e.g. in hydrogeological modelling of aquifers, due to their role in subsurface water movements. The description of small scale fracture systems is however a challenging task. In the test area (Kádárta, Bakony Mts, Hungary), two methods proved to be applicable to get reasonable information about the fractures: Electrical Resistivity Tomography (ERT) and Pricking-Probe (PriP). PriP is a simple mechani-cal tool which has been successfully applied in archaeological investigations. ERT results demonstrated its applicability in this small scale fracture study. PriP proved to be a good verification tool both for fracture zone mapping and detecting fractures, but in certain areas, it produced different results than the ERT. The applicability of this method has therefore to be tested yet, although its problems most probably origin from human activity which reorganises the near-surface debris distribution. In the test site, both methods displayed fracture zones including a very characteristic one and a number of individual fractures and determined their characteristic distance and significance. Both methods prove to be able to produce hydrogeologically important parameters even individually, but their simultaneous application is recommended to decrease


Introduction
35% of the land surface in Europe is covered by karst according to Karst in Europe COST 65 [1] and 25% of the global population is supplied by drinking water from karst aquifers [2]. These facts underline the importance of studying karst systems. Reference [3] demonstrated the importance of fractures in the development of a classical fracture-dominated karst aquifer. Several hydrogeological conceptual models have been developed for the characterisation of karst systems [4] [5] [6]. These models aim at qualitatively describing the hydrodynamic functioning of karst systems. The conceptual model produced by [7] and [8] provides a quantitative characterisation of karst and fractured systems. In this model, the spatial frequency of karst conduits is one of the crucial parameters influencing the hydraulic functioning of a karst or fractured system.
Understanding the hydraulic behaviour of karst systems is important for water research assessment, contamination risk assessment, vulnerability assessment, flood prediction, and speleological studies (e.g. [9] [10]).
Fractures are also important in engineering and geotechnical practice. They affect the stability of engineered structures and excavations [11]. Sinkholes, which develop along conductive underground features, are often responsible for large scale damages in artificial structures, and represent significant engineering issues.
Hydraulically, fractures behave as conductive features; however, in many cases they represent significant hydraulic barriers perpendicular to groundwater flow. Therefore, the identification, localisation and characterisation of fractures are crucial in studying karst and fractured systems.
At the catchment scale, it is possible to identify conduit locations e.g. on basis of sinkhole mapping [12] [13] or by geophysical methods. Most effective geophysical methods for identifying (individual) fractures and/or fracture zones (a dense set of fractures) include VLF-EM (e.g. [14]), VLF-R [15], VLF-EM-gradient [16], RMT [17], EM-34 [18], Electric Resistivity Tomography [19] [20] or geoelectric null-arrays [21] [22]. The resolution of these methods with the exclusion of the geoelectric ones is however smaller than it was required in the given study, where fractures were expected to be in even less than 4 -5 m distance.
In small scale, geotechnical tools would be perfect for fracture mapping, but they provide only point-like information. These methods are expensive and their application is strongly limited by field conditions, such as topography, artificial constructions, landslide risk or vegetation, which make the access to the study area difficult or even impossible. The Pressure-Probe method [23] which is a simplified version of the geotechnical instruments and which avoids all their aforementioned deficiencies may be an economic solution for such problems.
Reference [24] could detect fractures by GPR, but with only 5 m resolution.
The investigations carried out by [25] had at the same time very good resolution but they were carried out on a quarry wall, due to that the plateau above the cliff is covered with a conductive weathered layer, which drastically reduces the Journal of Geoscience and Environment Protection penetration depth of the GPR method. In the study by [26] ground-penetrating radar and frequency-domain electromagnetic induction methods proved to be capable to detect discrete fracture and conduit features. In Kádárta area GPR measurements proved to be unsuccessful.
Reference [27] [32]. The PriP was successfully applied previously in archaeological investigations [33] to localise structures such as walls or flooring at shallow depth. The first geological application of the PriP displayed the structure of the study area, which correlated perfectly with its main structural directions [34].

Site Description
The small village of Kádárta lies approximately 1 km northeast from the city of  the drinking water distribution network. The study of [35] and [36] identified diffuse contamination, and concluded that nitrate originated from the application of fertilisers throughout the Veszprém plateau.

Geological Settings
The geological environment of the study area belongs to the Transdanubian Range which itself most probably is a result of the NW-SE compression of alpic nappes in the Cretaceous period and the subsequent extrusion of the Alcapa terrain from the late Oligocene to the Miocene. This entire mountain range can thus be described as an allochtonous structural element [37]. The Transdanubian Range as a whole can be interpreted as a series of succeeding thrust sheets.
The eastern part of the Veszprem plateau consists of NE-SW tending strips of upper-Triassic calcareous and clastic sediments, dipping 15˚ -30˚ to the NW ( Figure 2). The recurrence of these strips is a result of thrust faulting along two significant reverse fault lines in this area.
In a late eoalpian compression, N-S oriented strike-slip and normal faults developed, resulting in the presence of multiple faults [38]. The Kadarta springs are located along one of these traverse faults, which are indicated by dry valleys. The  large plateau. This rock body mainly consists of highly fractured Triassic platform carbonates which prograded as a slope and toe-of-slope facies, and folded later in the Cretaceous due to NW-SE compression. The dimensions of this particular body of the BDF are approximately ten kilometres in length, two kilometres in width on the surface and could be 1000 metres deep at its thickest. Fracture zones in dolomite are profoundly significant in a hydrological respect, because of their high hydraulic conductivity mostly present in the so called "damage zone" of the ~10 -15 m wide fractured zones [40].
The study area is at a structural boundary, where the plateau rises due to the Veszprém Thrust, a fault formed in the Cretaceous deformation phase, crossing the field in its southern part and resulting in high fracture density in many places. The exact interpretation of this structural element, however, is still unclear. It is either usually described as a thrust or a normal fault, as well as new theoretical models suggesting sections of it being an oblique ramp [41].
Dolomite surface is covered by 5 -7 m thick quaternary loess in the SW, which is absent in the vicinity of the springs. The thin (10 -20 cm) poor quality soils directly overlie the dolomite surface in this area.

Hydrogeological Settings
The catchment area of the springs is about 20 km 2 . This area-as mentioned From a hydrogeological point of view, the role of this line is that it separates the hydraulically conductive Triassic BDF from aquicludes like the Veszprém Shale Formation, overlain by Iszkahegy Limestone, which form successive strata as part of a thrust sheet. This means that the Veszprém thrust itself can be determined as structural boundary between the nappes.
A large valley running into the ramp caused by the thrust can also be seen inside the area. The geological map of [42] indicates several parallel conductive fracture zones (Figure 2), corresponding to this valley well and other parallel trenches nearby.

Position of the Measuring Profiles
Both the ERT and PriP measurements were conducted on 4 parallel profiles (labelled P1 to P4 in Figure 3(a), Figure 3(b)). The direction of the parallel profiles was nearly N-S and they were adapted to the shape of the fenced area. The position of the profiles is presented also on the topographical map of the site (Figure 3(c)).

Electrical Resistivity Tomography (ERT)
The electrical Resistivity Tomography (ERT) is one of the most often applied geophysical techniques for shallow subsurface investigations among others due to the fact that the electric resistivity of the rocks varies in a very wide range enabling the separation of different rocks. Since electrical resistivity depends strongly also on the water content of a rock, this method can be very successfully applied also in hydrogeological studies [43].
In this study Wenner-Schlumberger (W-S) configuration has been used. Reference [44] stated that the W-S array is the most sensitive configuration to detect changes in vertical resistivity and more sensitive than some other arrays (such as the Wenner) to the horizontal resistivity changes. Its great number of data points and extensive horizontal coverage [29] also justified its application. Its robustness may be also very important in a variable environment.  In the field measurement a 72 electrode Syscal Pro Standard & Switch system was used with 0.5 m electrode spacing. This configuration is able to give an image up to 7.2 m depth. 1010 data points were used for the deep and 695 data points for the shallow W-S section.
The measured values have to be inverted to obtain a resistivity section which can be interpreted for hydrogeological purposes. The inversion was done using Earthlmager 2D Version 2.1.7 [45]. In the resistivity inversion settings, the stop criteria were set with 3% RMS error and 5% error reduction, because 3% noise level was assumed in the field taking into account that the measurements have been carried out in a village. The inversion terminates on meeting one of the criteria in these settings.
In the inversion of the field data L1-norm did not produce better results than L2-norm not even in the shallow sections ( Figure 8, Figure 9) which aimed to detect fractures. For this reason, all data were inverted using the L2-norm. Damping factor 1 was found to be the best. The RMS value which describes the fitting of the measured data and those calculated from the inverted model proved to be reasonable for all inversions without removing data.
To be able to detect fractures by the ERT method their electric resistivity values have to be different from that of the host rock. If the cracks in a dolomitic host rock are filled with clay or water, this criteria is satisfied, since their resistivity is less than 20 Ω•m contrary to the several thousands of Ω•m of the dolomite. In case of air-filled fractures the resistivity is very large. This strong variability of fracture resistivity can be confusing in the interpretation especially if the fractures are close to each other. Although water saturation could be estimated using factor analysis of engineering geophysical sounding data [46] we neglected it Journal of Geoscience and Environment Protection both due to economical and ecological reasons. Measurements were undertaken following a rainy period to assure that the fractures were still almost completely water filled that is they were conductive.

Pricking Probe (PriP)
The Pricking Probe method was earlier only used for archaeological exploration purposes [33]. Its principle is demonstrated in Figure 4. A T-shape metal rod with a sharp peak is pushed into the soil into a given depth equidistantly along a profile. If the rod cannot reach the given depth due to that it sticks in a rock a k value 1, if it is able to reach it a value 0 will be assigned to the given position.
Due to that samples are only taken at given locations there is certain randomness in the results. To decrease it a running average of a certain number of consecutive measurements used to be displayed integrating in this way the effect of the volume between the first and last measuring points. 5 -7 values seem to be practical to take into account in this process. The values determined in this way are called k 5 and k 7 and they are in the domain 0 to 1.    bedrock (denoted by thick green curve). The average size and the volume of the debris of the successive, equally thick "layers" approaching the surface are expected to be smaller and smaller due to the increasing weathering. With (decreasing debris volume, that is) increasing debris-free volume the probability that the Pressure Probe penetrates into the soil increases. Consequently k 5 or k 7 decreases with increasing distance from the bedrock.
Dissolution is more intensive close to a fracture because it serves as water flow path. Therefore also in the vicinity of a fracture the debris-free volume and thus the penetration probability must be larger than elsewhere. These areas are usually substantially wider than the fractures. In this way PriP method points to the existence of a fracture in a zone wider than the width of the fracture, enabling the application of a relatively great sampling distance. It also means that even rather narrow fractures may be detectable due to this effect.
The same was the situation in the study by [23] where the Pressure Probe method was applied, which is very similar to the PriP. In that study to detect fractures in loess it was definitely enough to apply a sampling distance three times the fracture width. Regarding the much larger consistency of the dolomite, the diameter of a valley due to a fracture may be much larger than the fracture width. A much weaker requirement is therefore enough for the sampling distance.
The diameter of the (often hidden) valley due to a fracture is the function in the first line the fracture width, dissolution capacity of the rock, dissolution time and average rainfall quantity. It is worth to note that sampling distance can be almost optionally decreased, arbitrarily improving the resolution of the PriP method. Our field measurements themselves verified (see later in the interpretation of profiles P1 and P2) that the resolution of the PriP is not weaker than that of the ERT method, if the circumstances are convenient.
"Measured" k and k 5 values are presented in Figure 5(b). Due to the randomness also at x = 0.5 m k = 0 value was taken, even if its probability is rather small. In zone x = 1 -1.2 m two k = 0 values were taken because the rode penetrates here into intensively weathered rock. The same is the situation in the vicinity of the fracture. k 5 values were then calculated (presented by triangles in Figure 5(b)) and their connecting line was displayed in Figure 5 ity. If at least one fourth, but less than three forth of the values are 1, the method is most likely well applicable. Not far from the study area the 0.3 m penetration depth was completely useless, because all measured values were 1 that is it was not possible to stick the rod into 0.3 m depth anywhere. Using however 0.1 m penetration depth the method proved to be applicable. In the given study area the standard parameters were used since they proved to be perfect.
We assumed the existence of fractures, where k 5 : 1. was below 0.2; 2. had a strong local minimum; 3. had a strong contrast in adjacent zones (see e.g. zones a, b and c in Figure 8(b)). In situations 1 and 2 the probability that the probe hits rocks is smaller due to that they are filled with fine sediments. In case 3 the alteration must occur due to sharp change in the depth of the bedrock that is by a fault, which also refers to the existence of a fracture. It means that all of these features most likely correspond to fractures.
The application of the PriP method is favourable because its application is very simple and it can be used even among the worst field conditions (extreme topography, weather). It is able to provide information also about the edges of the study area which may not be seen by ERT.

Major Sources of ERT and PriP Data Errors
Theoretically both methods should be able to detect fracture zones and even individual fractures. ERT measurements may produce convenient results if the electrical resistivity of the filling material of the fractures has a sufficiently large contrast with that of the dolomite and the distance of these features is sufficiently large in comparison to their depth. To get acceptable PriP results its sampling rate has to be small enough to get data from the fractures or their sufficiently eroded environment.
Even if the situation is theoretically convenient for the measurements, due to different noises the interpretation of the results may not be evident. Because resistivity is connected to fracture properties through water saturation it is recommended to carry out the measurements after a rainy period, when all fractures are more likely filled by water that is electrically more conductive than the host rock. In this way interpretation may be much easier. But even the results measured after a rainy period depend on many parameters (time of rainfall and its quantity before the measurements, water infiltration velocity, thickness of the cover, etc.) which also may make the ERT interpretation more complicated. The assumption, that the measurements are two-dimensional, are also often far to be valid. Among others fractures are rarely linear and there are present often also fractures orientated in other directions.
Due to the ambiguities of the ERT results it is recommended to verify them applying another method. The resolution of the VLF technique is weaker than required and GPR measurements could neither produce feasible results in the study site. Hence we decided to test the PriP method. It was its first verified geological application.
PriP data are affected by noise most often due to plants with thick roots which are also able to deny the probe to get into the required penetration depth. Since such plants can however be seen it is not difficult to treat the errors they produce. In our case there were not any such plants thus this kind of problem was irrelevant. Hollows of animal origin could also distort PriP results, but they did not occur in a large number and k averaging diminishes their effect rather well owing to their small extension similarly to the situation which was presented at x = 0.5 m in Figure 5. Human activity can however severely influence the PriP results by reorganizing the near-surface debris distribution. Building operation or agricultural activity may be the main causes of such distortions. In the study area artificial disturbances could only origin from building operations in the eastern part of the measure area. Their remnants may still be present in the area.
Positioning errors are misleading for both methods. It may not be as a serious problem measuring with one of the methods, but it is, if the aim is comparing the results of both techniques, especially if fracture density is high. In this case it is difficult to know whether a smaller shift between the anomalies of both methods is due to positioning errors or due to any other reason.

ERT Results 1: Overview Image
First of all the ERT sections are shown which served as a basis for all interpretations. P1 ERT section in Figure 6 does not display soil. It must be very thin. In  The decreasing resistivity towards P4 is unambiguous. It probably refers to the increasing sediment and/or water content which may occur due to higher fracturing. The most remarkable feature in the map is the about W-E oriented small-resistivity "trench" which starts from x = 12 -20 m in P1. It is interpreted to indicate zones with the highest water content corresponding to the most fractured part of the area. The direction of the trench is almost the same as that of the large fault which is 20 m right from this structure (Figure 3(a)).
It was shown that ERT can give a good image about the fracture zones. Now it will be discussed whether it is able to give a more precise image that is it can also detect individual fractures. This ability of the ERT will be compared to the one of the PriP method.

ERT Results 2: Detailed Image and Its Comparison with the Pricking Probe Results
We are going to study the shallowest ERT values in about 0.5 m depth (marked  Figure 8(a)), which are close to the investigation depth of the PriP. It enables the comparison of the results obtained by both methods. Resistivity variation must be due to the changes in the weathering level of the rock regarding that the study site uniformly consists of dolomite. More weathered dolomite has smaller resistivity due to the higher sediment and/or water content.
Three longer sections denoted a, b and c can be seen in Figure 8(a), whose resistivity values are remarkably different. As it has already been discussed zone a describes the highly fractured zone. Zone c must be a very compact dolomite, while in zone b the dolomite is moderately fractured. Narrow low resistivity zones, which are marked by ellipses, must be intensely dissolved, most likely due to fractures. Three such fractures are seen in Figure 8(a). The presented three fracture zones and the three fractures are the principal features visible in the resistivity section.
The PriP profile presents the same three zones as the ERT one. The average k level is remarkably different in these zones: it is 0.08, 0.25 and 0.92 in zones a, b and c, respectively, presenting the decreasing fracturing. The features which are supposed to be fractures are well seen also in the PriP profile. k provides 0 value in these locations (feature 1 and 2) or a strong local minima (feature 5). All of these features are at the same locations as the ones denoted by the ERT section.
Also the two significant resistivity maxima appear almost at the same positions where the PriP ones (red arrows). Two remarkable k minima (features 3 and 4) have at the same time no pair in the ERT section. Regarding, however, that they are at both ends of the highly resistive zone a, they are most probably linked also to tectonic features.
It can be concluded from Figure 8 that both methods provided principally the same results. They delineated three differently fractured zones and three significant fractures. It verifies that both methods would be able even individually

RMS=2.37%; L2=0.62 Journal of Geoscience and Environment Protection
Fracturing map can be constructed using the PriP results, too ( Figure 10).
The displayed quantity is 1 − k in percent. Where there are no fractures k is 1 (fracturing is 0%), while where the rod is going into the soil without hitting a rock (that is there are not unweathered rock matrix elements) k is 0 (fracturing is 100%). The larger is the proportion of the unweathered rock matrix elements in a given volume the smaller is the probability that the rod can penetrate into the given depth, the larger is k and the smaller is the fracturing.
The PriP fracturing map ( Figure 10) can be divided into three areas according to the fracturing level. Where there are the smallest values (zone I), the rock unit is very solid. Where are the largest values (zone III), the rock unit is strongly fractured. Zone II corresponds to a transition zone. Zone IIIa corresponds perfectly to the fault zone shown by the resistivity section in Figure 6 and Figure 8.

Conclusions
The aim of this study was to obtain information on the tectonic characteristics such as fracture patterns in a dolomite aquifer. Such information is indispensable e.g. in hydrogeological modelling. Beside the application of the Electrical Resistivity Tomography (ERT), the Pricking-Probe method has been tested, to see its applicability as an independent verification tool.
Beside of that, ERT produces information from greater depths supporting the interpretation of near-surface characteristics and the geology of the area, it can detect fractures, especially if the measurements are carried out following a rainy period and the distance between the fractures is large, in comparison with their depth. ERT results can also be used to construct porosity map, which is closely related to the fracturing one.
The PriP method is applicable to provide information about the subsurface conditions, when alterations manifest also near the surface. Since with increasing, fracturing the probe penetrates more likely into the desired depth, PriP is able to provide information about fracturing. Individual fractures can be regarded as very narrow zones with high fracturing. Since fractures are much wider at their top because the intensive dissolutions around them even narrow fractures are detectable with a reasonable sampling rate. Alterations in the k level may refer to faults.
Along one of the investigated profiles (P1), both methods provided principally the same results. They delineated three differently fractured zones and three significant fractures proving that both methods would be able to display such fea- The study demonstrates that the presented methods are able to describe the fracture patterns of an area. They are able to characterise both fracture zones and fractures. Such information is very useful e.g. for hydrogeologists and engineers.
In the study area, it was concluded that the surface of the unweathered dolomite is deepening both eastwards and northwards, similarly to the surface topography. A wide fracture zone was discovered on basis of the fracturing maps by both methods. This zone would be ideal for extracting water. A number of individual fractures have been localised in the west, in the moderately fractured areas at an average distance of about 3 m. In the eastern part of the site, in the highly fractured areas, there are more uncertainties, partly maybe due to the supposedly smaller 1 -1.5 m fracture distance. This knowledge is principal in building hydrogeological models for the area.
ERT seems to be effective in describing the fracture pattern in spite of the uncertainties. To avoid the misinterpretation, further measurements with an independent method are recommended. Although its applicability and limitations have still to be studied PriP proved to be convenient for this aim. Furthermore, it may be the only applicable tool in small scale fracture studies close to the edges of the study site or in extreme field conditions, such as e.g. a very rugged terrain. The studied methods even individually can provide important information for hydrogeologists and engineers. Their joint application may even be more fruitful.