The Use of Electrical Resistivity Tomography to Investigate Basaltic Lava Tunnel Based on the Case Study of Al-Badia Cave in Jordan 1

Electrical Resistivity Tomography (ERT) was employed to conduct a geoelectrical survey near the AlBadia lava tunnel located close to the Al-Bishyrria Village in Jordan. The technique enabled the mapping of the subsurface tunnel extension and description of its inner structure. To assess the quality of data and resistivity models, Schlumberger and Reciprocal Schlumberger electrode configurations were used to produce eight ERT profiles. As revealed by the examination of received potential, the implemented configurations exhibited a strong signal, producing an approximated reciprocal error of up to 6%. The findings of ERT models showed that the lava tunnel had a clearly outlined structure with an elliptical to rectangular shape. The modelled resistivity of the lava tunnel was obtained in proximity to 1000 Ω-m, with a better characterization being possible at resistivity exceeding 8000 Ω-m in 200 Ω-m of Fahda Vesicular Basalt medium. An exploration depth of 50 m revealed that the lava tunnel was 10 m deep and 5 m in diameter on the average. Furthermore, potential means of groundwater recharging were reported by the simultaneous detection of a number of resistivity anomalies of less than 50 Ω-m and lava tunnel. In addition, the lava tunnel was observed to extend and ramify beyond the area under investigation, indicating at the potential existence of multiple lava tunnel extensions in both the investigation area and in the basaltic flows, which could have adverse implications for future urban projects.


Introduction
Located in north-eastern Jordan, the intra-plate Harrat Al-Shaam Basalt of Neogene-Quaternary period is believed to be the northern extension of the north Arabian province . Spanning a distance of over 3,000 km, it spreads north-south in alignment to the Red Sea axis, from Syria through Jordan to Saudi Arabia ( Figure 1a) . Jordan com-prises 11,000 km 2 of around 46,000 km 2 of overall surface area (Ibrahim, 1997), earning it the status of the volcanic field of the greatest size in the Arabian Plate (Ibrahim and Al-Malabeh, 2006). The basaltic volcanic province of North Arabia has been the focus of extensive researches (e.g. Lartet, 1869;Doss, 1886;Van Den Boom and Sawwan, 1966;Bender, 1968Bender, , 1974Abed et al., 1985;Al-Malabeh, 1989, 1994Ibrahim, 1993aIbrahim, , 1996aIbrahim, , 1996bHall, 1995, 1996). The Natural Resource Authority (NRA) has comprehensively mapped the geology of this province . The Harrat Al-Shaam basaltic super group is a classic intra-plate continental basaltic flow of alkali olivine basalt comprising the basaltic provinces and are classified into five groups (Ibrahim, 1993a). The K-Ar dating method produced a date of between 13.7 Ma to less than 0.5 Ma for the volcanic activity in northeastern Jordan (Barberi et al., 1979;Moffat, 1988;Tarawneh et al., 2000;Illani et al., 2001).
In northeastern Jordan, basaltic lava flows are arranged in sedimentary successions in an unconformable way. Within the studied area, the direction of these flows is approximately to the south and southeast, as shown in Figure 2a. Basaltic  Kempe et al., 2006) superimposed on a Google Earth map; c. Schematic representation of the inner configuration of the Al-Badia lava tunnel (modified after Kempe et al., 2006); d. General view of Al-Badia Basaltic Cave; e. Rainstorm flood at Al-Badia basaltic cave (photo taken from Al-Adamat, 2018). lava is not homogeneously thick; for example in Syria, it has a thickness of about 1 km, but at the border between Jordan and Syria it is around 600 m thick, and its thickness decreases to the south, becoming interspersed with sediments (Gibbs, 1993). Geological field surveys conducted by Bender (1968) and Al-Dmour (1992) uncovered a number of basaltic lava flows, with a minimum of fourteen basaltic lava flows of different thickness being discovered in the north of the Azraq Basin by deep penetration drilling (NRA, 2008). In the area of the Harrat Al-Shaam, surveys and explorations of several basaltic lava caves carried out, have been adequately mapped and documented (Arsalan, 1974;Al-Malabeh, 2005;Kempe et al., 2012). Kempe et al. (2012) extensively investigated twenty such caves, then describing their locations and extensions and categorizing them into types in accordance with how they were formed and their inner structures, including lava tunnel (pyroduct), pressure ridges, which some of them are of unknown origins.
Lava tubes, lava tunnels, and volcanic structures in different parts of the world have been the focus of multiple geophysical studies. For instance, the lava tube in the northeastern Jordanian region of Umm Al-Quttain has been explored by Al-Oufi (2006) and Al-Oufi et al. (2012) with the two-dimensional techniques of induced polarization imaging (2D-IP). The extension of two lava tube faults and dikes in northeastern Jordan was investigated by Al-Oufi et al. (2008) based on a very low frequency (VLF) electromagnetic technique. Moreover, the techniques of ground penetration radar (GPR) and electrical resistivity imaging (ERI) were combined by Ortiz et al. (2007) to explore volcanic materials and features in Tenerife, Spain. GPR was also used by Miyamoto et al. (2003) to identify lava tubes in the Komoriana Cave at the Fuji Volcano, Japan. Detection of volcanic structures of superficial depth in a region of high potential value was conducted by Ortiz et al. (2014) within various geophysical prospecting techniques, such as microgravity, GPR, and electromagnetic induction (EMI). The lava tunnel of Manjanggul on Cheju Island was investigated by Kwon et al. (1998) The purpose of the present study was to explore the shallow structures of the Al-Badia lava tunnel in the Al-Bishriyya region, Jordan based on electrical resistivity tomography (ERT). A comprehensive assessment of the efficiency of Figure 2. a. Simplified geological map of the studied area (Modified after Ibrahim, 1997); b. Columnar log description for BY (Bishriyya Group) . two-dimensional resistivity data obtained in the researched area was undertaken with regard to data quality and resistivity models. Furthermore, to determine whether the underground lava tunnel structure could serve as a source of groundwater recharge in an area of high aridity, the extensions of the lava tunnel were investigated through the resistivity models.

Research Area
To ascertain the usefulness of ERT for exploration of known tunnel subsurface extensions, the present study focused on the case study of the Al-Badia Cave. Otherwise, referred to as the Al-Huwinit or Beer Al-Hamam cave, this cave received its name from Al-Malabeh (2005), and it is considered to be a lava tunnel basaltic cave . Based on the WGS 1984 coordinate system, the coordinates of the cave are 32°07'57.06"N and 36°49'25.14"E, which is in the proximity of the Al-Bishriyya Village ( Figure  1). Kempe et al. (2006) conducted a comprehensive survey of the cave with the purpose of mapping its inner structural features. The subsurface extension and the inner features of the Al-Badia lava tunnel are represented in a schematic form in Figures 1b and 1c (Kempe et al., 2006).
The Al-Badia lava tunnel has an overall volume of 20,000 m 3 (Shawaqfah et al., 2014), 15 m wide and of 8 m high. Its access point is a roof collapse ( Figure 1d) with a depth of 5 m. Its location on the path of an important wadi means that, during rainfall, it is gradually filled with fine sediments and loess transported by the running water (Figure 1e), that are thick enough to block its extension . A survey and mapping were conducted on just 445 m of the tunnel , but its subsurface extensions are anticipated to be larger. Groundwater recharge depends greatly on the high volumes of surface water stored by the lava tunnel during floods ( Figure 1e) (Al-Amoush, 2010;Shawaqfah et al., 2014). A simulation of the groundwater was undertaken by Shawaqfah et al. (2014), revealing that the storage of surface water by the Al-Badia lava tunnel led to an increase of 0.2 -0.5 m in the level of local groundwater.
As suggested by Al-Malabeh (2005), the lava tunnel may have potential uses in the context of environmental tourism.

Geological Characteristics of the Research Area
The geology of the studied area is mapped in a basic form as shown in Figure 2a (Ibrahim, 1997). The area is composed of several geological formations (Ibrahim, 1997).
Dating to the Pleistocene-Holocene, the Bishriyya Group (BY) consisting of the latest volcanic activity in northeastern Jordan, is associated with the majority of B6 basalt classified by Van Den Boom and Sawwan (1966) and Bender (1974). According to the truncation of flow lines that showed distinct ages, the group is separated into two formations, namely, the upper division of Fahda Vesicular Basalt Formation (FA) and the lower division of Wadi Mansif Basalt Formation (WMF). The WMF is not exposed in the studied area as indicated on the map (Figure 2a). On the north side, BY projects from volcanic vents is flooded along earlier wadis. A columnar log description of BY in the researched area is provided in Figure 2b . From a morphological perspective, BY consists of fresh blocky "aa" and some "pahoehoe" flows. The direction of the basalt lava flow is indicated by a number of pressure ridges, which present shrinkage joints with a width of up to 50 cm and length of 6 m that are usually considered to be the products of cooling of basaltic lava. Furthermore, polygonal columnar jointing is another characteristic feature, and records of preserved lava tubes have been made at numerous sites within this group in Umm El-Quttain region. The phenocrysts making up BY are melanocratic grey, with fine grains, and colour that varies between green and yellowish olivine . The K-Ar isotope dating technique produced a date of between 1.45 Ma to less than 0.1 Ma for the FA formation (Baraberi et al., 1979 andMoffat, 1988). However, Tarawneh et al. (2000) obtained an age range that differed somewhat, between 3.5 Ma to less than 0.5 Ma.

I J O G
The Use of Electrical Resistivity Tomography to Investigate Basaltic Lava Tunnel Based on the Case Study of Al-Badia Cave in Jordan (H. Al-Amoush and J.A. Rajab)

165
Every 2D-ERT profile was 235 m long, and for all conducted profiles, Schlumberger (S) and Reciprocal Schlumberger (RS) configurations were employed. RS indicates where the injection current electrodes are relatively located internally towards the middle of the array, while the position of receiving potential electrodes is external to electrical current electrodes. By contrast to the standard Schlumberger array, this configuration exchanged current and potential pairs, so that the configuration factor K stayed unchanged (Figure 3b).
At the studied area ( Figure 1d, 3c), lava tunnel outcrop appears as an air-filled cavity structure overlain by superficial clay deposit layers. This model makes the ground condition to be highly conductive over highly resistive cavity structure. The Schlumberger (S) configuration was applied, because it was characterized by good depth of penetration and resolution where the applied current and received signals are of high signal-tonoise ratio (S/N) (Loke, 2013). Moreover, the S

Strategy of Field Survey and Data Collection
Two-dimensional resistivity data were obtained at eight sites in the area of Al-Badia tunnel cave with the multi-electrode 2D-ERT technique ( Figure  3). The strategy of field data collection was geared towards attaining a representative resistivity signature of subsurface tunnel structures by executing a number of profiles throughout or near the lava tunnel. Apart from the 2D-ERT profiles of ERT04 and ERT06, which were oriented in east-west direction, the other five profiles of ERT01, ERT02, ERT03, ERT05, and ERT07 were oriented mainly in a north-south to northeast -southwest directions ( Figure 3a). Furthermore, the profiles carried out from potential lava tunnel extensions that were not known were ERT06 and ERT07 ( Figure 3a). Resistivity was measured by Syscal Junior Switch instrument ( Figure 3c) via forty-eight electrodes which were arranged at 5 m electrode spacing. configuration has a slightly better spatial resolution than other common resistivity configurations (i.e., Dipole-Dipole and Wenner configurations) which make it becomes sensitive to resolve high-resistivity structure at comparable models. In addition, RS configuration tends to quantify random noise at larger values as it picks up more noises, being useful to assess such noises at higher levels (Dahlin and Zhou, 2004). The choice of S and RS model-based configurations implemented at this study can be compared and assessed by the synthetic model composed of high-resistive buried channel embedded in conductive structures (similar to our lava tunnel model) presented by Dahlin and Zhou (2004). Their resistivity modelbased S configuration shows a good resolution of buried channel at low noise contamination where S/N is quite high. The derived model also has accurately resolved and mapped the overburden layer of moderate conductivity.
The ERT is designed in such a way that makes it possible to record 565 quadripoles arranged over 21 levels, and has an apparent penetration depth of around 50 m. Between the first and 21 st resistivity level, K varies from 31.42 m to around 2061.67 m.

Quality of Data and Evaluation of Errors
The noise of the resistivity measurements can be approximated by determining a quality factor based on an observed potential error (i.e. reciprocity error), derived from the normal measurements attained from S configuration and the reciprocal measurements attained from RS configuration. To obtain the 1% standard deviation of the calculated potential, the configuration of the instrument system is set on acquisition of resistivity measurements (i.e. quadripoles) at six stacking, current pulse length of 1.0 s, and instrument voltage of 800 V.
Ground resistance of less than 9 kΩ was exhibited by the obtained resistivity data, with around 200 mA maximum injected current and received potential attained at the highest value of 55 mV driven by instrument power up to 400 V.
Apart from the quadripoles, which displayed significant disturbances close to the first level equivalent to an apparent depth of 2.85 m, a less than 6% stacking error arose in the conducted survey. The fact that the superficial conductive clay layer and the resistive layer underneath (Fahda Vesicular Basalt formation) came into contact might explain the disturbances.
The stacking error is typically lower compared to the reciprocal error (Binely et al., 1995). Noise content and magnitude in resistivity pseudo-section can be better estimated based on reciprocal error, whereas the data quality afforded by the low stacking error is poor (Zhou and Dahlin, 2003). A direct correlation may be established between a number of intrinsic errors and various sources; to give an example, background noises of an ambient and anthropogenic nature, physical instrument dysfunction, and high contact resistance of electrodes used in measurements, which supply a source of excess and non-linear potential dominate imaging results (Ramirez et al., 1999;Zhou and Dahlin, 2003;Wilkinson et al., 2008).
As shown in Figure 4, the reciprocal errors of the eight ERT profiles take the form of a logarithmic plot of absolute relative potential errors, the calculation of which is based on the normal and reciprocal measured potential. The plot comprises statistical parameters such as mean M and fitting regression line as well; for instance, reciprocal error for the illustrated data was defined and quantified based on standard deviation SD.
By examining the relative potential errors, it was found that outliers have a suboptimal performance in the part of the plot (Figure 4) associated with measured potential of less than 1 mV, while other outliers were related to a relative potential error exceeding 15%. On the whole, by analyzing the potential errors of the eight ERT profiles, it could be observed that the employed configurations had suitable signal strength, since the majority of measured data were in the region of 10-15 mV of measured potential error.
The reduction in the relative potential error due to the rise in the measured potential up to 15 mV resulted in a negative trend in the regression lines. The SD of the relative potential errors (%) in the range 2.9 -6% was used for representation of the attained reciprocal error. For each ERT profile, relative resistivity pseudo-sections were created in order to enable understanding of how the relatively measured potentials and their sequential apparent resistivity values were spatially assessed ( Figure 5). The results show that the majority of plotted data were observed to be within ±4% of resistivity variations and were congruent with around 90% of reciprocal data. Several high-positive or highnegative noisy data occurred in localized zones throughout and at the side of pseudo-sections (i.e. ERT03 and ERT05). In addition, a correlation existed between numerous noisy data and greater depths, with measurement of noises at the longer potential electrode separation or large K values (i.e. ERT02, ERT06, and ERT08). Random noisy data dominated at every level of measured data, as exhibited by some relative resistivity pseudosections (i.e. ERT01, ERT06, and ERT07).

Inversion of Resistivity Data
As noted by Kempe et al. (2006), the manner in which basaltic lava was geologically constituted and how it was structured underground revealed that a lava tunnel was an inhomogeneous three-dimensional feature with regard to depth, form, and the occurrence of bulk spaces that loaded with fine sediments. Non-random sources produce the noises in the outliers present in the resistivity datasets. Given that the reciprocal errors demonstrated by all the resistivity data were less than 10%, the impact of outliers on the resistivity measurements was minimized through the application of the blocky or robust inversion routine. The purpose of the robust resistivity model accompanying the L 1 -norm inversion is to reduce the absolute difference between the measured and calculated apparent resistivity values as much as possible (Loke et al., 2003). RES2DINV software permitted inversion of the resistivity data (Loke, 2002), and the quality indicator of inverted models was taken to be the root mean squared error value (RMSE %). The most common usage of robust inversion is in detection and refinement of sharp boundary over pronounced resistivity contrast (Loke, 2002). Furthermore, smoothness-constrained least square inversion, which is characterized by the fact that the L 2 -norm inversion displays sensitivity to outliers (de Groot-Hedlin and Constable, 1990) was employed as well in this study. A relatively higher RMSE value was obtained from this inversion, generating comparable resistivity image of major signatures, despite the slight change in the resistivity scale and less resolved areas of sharp resistivity contrast (Figure 6a).
To give an example, the implication of the ERT01 pseudo-resistivity section conducted in the proximity of exposed lava tunnel is that resistivity distributed with a clear interface among various regions at distinct resistivity values and the resistivity value in every region is smooth and constant (Figure 6b). An anomaly with an ellipsoidal form and high resistivity (~1,000 Ω-m) was found to develop between the horizontal distance of 90 m and 110 m at approximately 5 m depth, according to the robust resistivity model. Furthermore, numerous sites were found to have conductive features (~50 Ω-m) close to the surface, which could have been linked to similar deep conductive features under the anomaly of high resistivity. The robust resistivity model provides a satisfactory explanation for this. It is worth noting that due to the three-dimensional effect that lava tunnel implies, such setting can give rise to artifacts in the resolved model as a two-dimensional forward scheme used for modelling of the three-dimensional structure (Griffiths and Barker, 1993).

Interpretation of Data and Results
The inverted resistivity data took seven iterations to achieve convergence of the resistivity models of the ERT profiles, and the associated RMSE values of less than 7.7% were constrained and retained their prominence with the reciprocal data error, considering that the proportion of data filtering was around 8% (Figure 7). Since logarithmic variations in resistivity values are known to occur, a fixed logarithmic colour scale was employed to present each resistivity model, thus enhancing clarity and enabling comparative analysis between the models. There was consistency between the ERT01 model and the inner . a. Resistivity model associated with smoothness-constrained least square inversion at ERT01, with 9.28% RMSE for L 2 -norm inversion according to the plot of calculated and measured resistivity data. The acquisition of the inverted models occurred after 7.8% cut-off noisy data; b. Resistivity model associated with the robust inversion at ERT01, with 6.7% RMSE for L 1 -norm inversion, according to the plot of calculated and measured resistivity data.
structure at the collapsed ceiling section of the lava tunnel. The air-filled lava tunnel with a diameter of 5 m representing anomaly "A" was defined by ~1000 Ω-m and positioned 5 m deep in the medium of ~200 Ω-m of the Fahda Vesicular Basalt Formation. The floor of the lava tunnel is flat in this area, with basaltic contraction cracks (Figure 1c) (Kempe et al., 2006) and the forma-  Figure 7. Resistivity models associated with the resistivity profiles; according to the underground map, the known lava tunnel is indicated by anomaly "A" (Kempe et al., 2006), while anomalies "B", "C", "D", and "F" may represent extensions of "A" or new lava tunnels. The numbers next to the anomalies (i.e. 7, 11, 14/15, 16, 27, 29/30) indicate how the anomalies of the ERT profiles and their positions are spatially coordinated and correlated in plan view and longitudinal cross-section (Figure 1c).
tion of an anomaly of less than 50 Ω-m resistivity to a depth of approximately 50 m. The subsurface lava tunnel network (Figure 1c) was followed to determine how the resistivity models were affected by the structure of the lava tunnel. Moreover, to provide additional elucidation, a number of features associated with the inner structure, like lava tunnel filling materials and dimensions, were used as well.
A resistivity value exceeding 3,000 Ω-m and even as high as 10,000 Ω-m was associated with the resistivity anomaly "A" denoting the lava tunnel and attained at ERT02, ERT03, ERT04, ERT05, and ERT08, while its depth was around 10 m. Furthermore, its diameter varies from 3 m at ERT08 to around 7 m at ERT03 and ERT04, suggesting a deep channel (Figure 1c and Figure 7).
Detected at ERT02 and ERT04 (Figure 7), the resistivity anomaly "B" exhibited a resistivity value exceeding 3,000 Ω-m reaching up to 10,000 Ω-m, with the subsurface tributary blocked by sediments being believed to represent a link to the primary lava tunnel (i.e. anomaly "A") ( Figure 8). The lava tunnel is likely to extent further southwards as suggested by the anomaly "D" detected at ERT06 (Figure 7), which was around 10 m deep and has a resistivity value of around 3000 Ω-m.  Figure 8. Inner structure of the Al-Badia tunnel (modified after Kempe et al., 2006) superimposed with the sites of ERT profiles and identified resistivity anomalies. Anomaly "A" with a high resistivity indicates the tunnel, while anomalies "B", "C", "D", "E" and "F" may suggest unknown tributaries of anomaly "A".

I J O G
The Use of Electrical Resistivity Tomography to Investigate Basaltic Lava Tunnel Based on the Case Study of Al-Badia Cave in Jordan (H. Al-Amoush and J.A. Rajab) At ERT07, anomaly "F" was detected, having a resistivity value of around 1,000 Ω-m and potentially representing a northward drift of anomaly "A". Moreover, the formation of the resistivity anomaly "E" with a depth of 10 m was revealed by ERT07, while the formation of the anomaly "C" with a depth of 5 m was revealed by ERT08. These two anomalies could be conjectured as additional proof that the lava tunnel extends further northwards (Figure 8).
The derived ERT models for lava tunnel in this study can be compared to the results of other geophysical methods (e.g. Al -Oufi et al. 2008). The Very Low Frequency -Electromagnetic (VLF-EM) inverted models presented by Al-Oufi et al. (2008) at Umm El-Quttein area, NE Jordan, suggest very long lava tunnel extends to more than 5 km. Possible lava tunnel divergence was recorded along the survey lines. Also, the quantitative modelling of VLF data have resolved lava tunnel with resistivity of over 2,500 Ω-m, and better determined in less resistive Fahda Vesicular Basalt medium. Furthermore, the subsurface extension of lava tunnels were delineated within the first 20 m depth at all places of conducted VLF profiles and associated of multiple conductive structures of filled-sediment joints and fractures.

Discussion
In northeastern Jordan, there are volcanic eruptions and basaltic lava flows dating to the Neogene-Quaternary, with associated lava tunnels and caves forming over an extensive area and recorded to a shallow depth. Consisting of the upper Fahda Vesicular Basalt Formation and the lower division of Wadi Mansif Basalt Formation, the Bishriyya Group includes most of the lava tunnels of limited depth. The characteristics of the two formations include both horizontal and vertical joint systems, vesicles, and fresh blocky "aa" lava.
This study sought to explore interconnected underground extensions and inner structure of the Al-Badia lava tunnel, and to this end, eight ERT profiles were performed. For the purposes of target imaging, the ERT configurations and measurement of resistivity data quality employed to describe both known and unknown lava tunnels demonstrates a satisfactory efficiency. Data of good quality were obtained through the resistivity datasets based on Schlumberger and reciprocal Schlumberger with equivalence between ± 4% of resistivity variations and 90% of reciprocal data ( Figure 5). The lava tunnel was resolved in an adequately formed structure based on a maximum 10% of filtered data and application of the robust inversion routine (Figure 7).
Besides the mechanisms underpinning the development of the lava tunnel, it was observed that the setting of the lava tunnel in the researched area had a minimal ground gradient (Kempe et al., 2006), which made it possible for the structure of the tunnel to be broadened (Calvari and Pinkerton, 1998;Kempe et al., 2006). The variation in the dimensions of the tunnel reflects this, with the width between 10 and 20 m, the diameter between 3 and 7 m, and the depth to the ceiling of the tunnel between 5 and 13 m.
According to the geography of the Al-Badia tunnel cave, the location of the main tube-fed "aa" lava flows is at the distal part of the foremost lava flow field system, as can be seen in Figure 2. This type of lava flow spreads more widely and faster compared to other kinds of lava flows due to the low viscosity, accelerated speed, and great flow thickness given by the lava tunnel. Consequently, considerable heat is lost on the lava flow surface owing to the created lava tubes and channels (Melnik, 2017). The conclusion drawn is that the formation of the lava tunnel under investigation occurred under high discharge flux, with the earliest lava flow being induced in a northwest direction and approximately 150 km away from the researched area. Over 0.5 km of the length of the lava tunnel with 5-m average diameter was revealed by the restricted survey grid.
Besides delineating known sites along the lava tunnel, namely, anomaly "A" (Figures 7 and  8), the ERT models employed revealed the formation of a secondary lava tunnel as well, resulting from shifts in the orientation of the lava flow. It could thus be implied that there was a decrease in the original lava flow rate over effusion rate due to which a secondary drainage passages developed, namely anomalies "B", "E", and "F" (Figures 7  and 8). Kerr et al. (2006) investigate the behaviour of lava spreading that forms a point source using a laboratory experiment and scaling theory at a constant flow rate on a wide uniform slope. The study attributed the lava tube channelization to the ground slope, pre-existing topography, lava viscosity, lava yield strength, surface solidification, and rate of propagation.
During the formation of an air-filled tunnel, the pressure of primary lava flow maintained constant consistency and strength, ultimately resulting in self-support of the inner geometrical structure, as suggested by the fact that within the survey scale, the collapse of the lava tunnel ceiling was not represented by the ERT models. The lava tunnel exhibited well-delineated rectangular to ellipsoidal forms, high resistivity contrast, and resistivity exceeding 1,000 Ω-m within 200 Ω-m of Fahda Vesicular Basalt formation. However, in a number of areas, it has a weak and highly-jointed floor, enabling formation of a preferential path for fine sediments and surface water penetration (i.e. ERT01, ERT04, ERT07, and ERT08). Shawaqfah et al. (2014) and Al-Amoush (2010) suggested that such structures may mediate groundwater recharge. Observation of this was made at different areas with the resistivity of less than 50 Ω-m below or near the lava tunnel floor (e.g. ERT01 and ERT08 in Figure 7). Hence, recovery of conductive anomalies occurred solely where the resistivity of the lava tunnel was minimal (i.e. ~1000 Ω-m). So far, various lava flows with overall thickness of 600 m have been discovered in northeastern Jordan (Gibbs, 1993). The developments of near-surface and deep lava tunnels may be furthered by the Bishriyya Basaltic Group.

Conclusion
Focusing on the Al-Badia lava tunnel, this study proves that ERT could effectively map the underground structure of a lava tunnel. Data error quantification is based on the reciprocal error, so the recovered resistivity models have a maximal quality due to the robust signal strength up to 50 m deep supplied by the Schlumberger and its reciprocal configurations.
The resistivity models reveales that the geometry of the lava tunnel is rectangular to ellipsoidal in form, 3 -7 m in diameter and 3 -13 m in depth. The resistivity of the air-filled lava tunnel also varies in the range of 1,000 -10,000 Ω-m. The lava tunnel was identified in ~200 Ω-m solidifying lava corresponding to the Fahada Vesicular Basalt Formation. Ellipsoidal structures with resistivity of less than 50 Ω-m are identified in numerous lava tunnel sites, especially sites where fine sediments and/ or water have leached through cracks in the basaltic floor, for instance in areas where the lava tunnel or nearby sites are invaded by conductive water and fine sediments. The ERT resistivity models show that known and unknown lava tunnels are directed to the northwest and south of the researched area.