Feasibility and limits of Electrical Resistivity Tomography to monitor water infiltration through karst medium during a rainy event. S. D. Carrière (1*), K. Chalikakis (1), C. Danquigny (1), R. Clément (2) , C. Emblanch (1) (1)
UAPV, UMR1114 EMMAH, F-84914 Avignon, France
(2)
IRSTEA, 1 rue Pierre-Gilles de Gennes CS 10030, 92761 Antony CEDEX, France
* Corresponding author :
[email protected]
Abstract The common hydrogeological concepts assume that water enters in karst media by preferential pathways. But it is difficult to identify these pathways, particularly if soil or scree covers the karst features. When and where does water enter in the hydrosystem? How fast? A unique large scale Electrical Resistivity Tomography (ERT) surface based time-lapse experiment was carried out during a typical Mediterranean autumn rainy episode (230mm of rain over 17 days). 120 ERT time-lapse sections were measured over the same profile during and after this event (30 days). The main goal was to evaluate efficiency and limits of the ERT to monitor water infiltration, under natural conditions. Apparent (directly measured) and inverted resistivity’s variation during the rainy event highlights some interesting zones. They could be interpreted as preferential pathways, where water dynamic seems quicker in term of moistening and drainage. Nevertheless, these results have to be interpreted reasonably because ERT does not provide enough precision to determine exact pathways geometry and functioning. In addition, forward modeling provided relevant data treatment limitations mainly for the deeper parts of the sections. 1. Introduction Identifying and locating potential preferential pathways such as conduits, faults or fractures is peculiarly difficult in karst areas. In addition, characterizing and locating water movements through these pathways is probably one of the most important and challenging task in order to improve enhanced understanding of karst water dynamics. Hydrogeologists have been developing several methods mainly related to natural and artificial tracing to access these water pathways. Geophysical tools could help identifying these potential pathways with a suitable methodology [e.g. 15, 27, 6, 22, 4]. However, even if such features were fully identified, effective water flows pathways would probably remain almost impossible to predict and monitor. In karst areas the link between structure and hydrodynamic functioning remains a bottleneck. This could probably explain also the limited use of physically-based gridded flow models for karst hydrodynamic modeling. Ground Penetrating Radar (GPR) is probably the most efficient geophysical technique to image near surface features under adequate surface conditions [eg. 2, 6, 4]. However GPR does not allow identifying if these features impact water dynamic. Electrical Resistivity Tomography (ERT) technique due to its robustness and reliability has been widely used in karst areas to identify karst features [e.g. 29, 3, 28]. In porous semi-homogenous media, ERT, due to its sensitivity to water content, is a commonly used geophysical technique, fast enough to follow up water infiltration during rainy events [e.g. 11, 7]. A unique large scale ERT surface based experiment was carried out during a typical Mediterranean autumn rainy episode (17 days) in the LSBB (Low Noise Underground Research Laboratory, Rustrel France) within the Fontaine de
Vaucluse karst hydrosystem (Fig.1). A total of 230mm of rain was registered and 120 ERT time-lapse sections were measured over the same profile during and after the event (30 days). The main goal was to evaluate efficiency and limits of the ERT to monitor water infiltration via previously recognized karst features [4], under natural conditions.
Fig.1 : Fontaine de Vaucluse basin located in France; (b): Test site located in Fontaine de Vaucluse basin (after[24]); (c): Aerial photo with main lineament detection and survey location (www.geoportail.gouv.fr); (d): Regional lithostratigraphic log ([21] (modified)).
2. Implantation/acquisition strategy and field constraints At the experimental site, the slope, the vegetation and the gravel cover induced a very important preparation work to clear the ERT chosen profile before measurements. In addition, the implantation of electrodes, mainly at limestone outcrops, was an additional difficult task; electrode holes were mechanically dug within the rock; saltwater and mud were used to ensure a good quality ground contact. Bad electrode-ground contact could induce artifacts during data acquisition. These artifacts are comparable to near surface inhomogeneities [25, 1] which could deteriorate the final inverted resistivity section [16]. The chosen ERT profile follows an East/West direction. This direction is perpendicular to the general slope, it is sub-perpendicular to one of the main faulting and lineament directions and it is the most heterogeneous direction, in term of apparent resistivity spatial distribution [4]. The acquisition system used for ERT time-lapse measurements is an ABEM Terrameter SAS 4000 [9] with 4 channels and 64 electrodes. The profile has 126m length and a 2m inter-electrode spacing. Electrodes exact position was measured with Real Time Kinematic (RTK) acquisition method every 2m using a differential GPS (TRIMBLE GPS 5800 with TSC), with an accuracy around 1cm for X and Y and 2cm for Z. The electrodes stayed in the field during the entire campaign. For the time-lapse measurements the gradient array was chosen for his robustness and rapidity [10]. This protocol totals 1360 measurement points. For each measurement point, the acquisition time was 0.1sec and the delay time was 0.2sec.
Each measurement point cycle spend around 1.8sec. To ensure data quality, during acquisition if a data point presented a repetition Root Mean Square (RMS)>1%, the measurement was repeated until five times. A 50Hz filter was also applied to reduce anthropogenic noise. During the 30 days campaign, an ERT time-lapse section was acquired every three hours during the rainy event (17 days) and every day until two weeks after the rainy event. 3. Data treatment and results 3.1. Classical processing During this monitoring campaign, a total of 120 ERT sections were acquired. When data acquisition quality was not satisfactory (repetition measurement of all the section RMS > 2%) the section was removed. Finally, a total of 106 ERT sections were kept to follow data treatment. Apparent resistivity values have been averaged for each section. This mean apparent resistivity decreases strongly during the rain event, from 1750 to 1050 Ω.m (Fig.2A). These variations do not seem related with temperature variations because air temperature remained stable during the campaign [5]. Thus, we can reasonably relate these resistivity variations with water content variation. Analysis of this mean apparent resistivity indicator allows selecting eight critical times step presented in Fig.2B and Fig.2C, before, during and after the rainy event. Several processing strategies and softwares were used for data treatment and analysis. A commonly used software, Res2Dinv [19, 20] and a research software package, including DC2DinvRes and BERT (Boundless Electrical Resistivity Tomography) [13, 26, 14]. These results were presented and discussed in details in [5]. The inversion results presented in Fig.2B were performed using DC2DinvRes. We used a standard time-lapse inversion following the approach proposed by Loke in [16]. First, the initial time step model was computed. Second, we used it as a reference model for the other time step. Finally, we compared the resulting calculated models (Fig.2C) using percentage change in calculated resistivity (Δρcalc), Eq1.
calc 100 0 1
[1]
Where ρ0 is the calculated resistivity at the first time step, ρT is the calculated resistivity at time step “n”. During the four first selected time-steps, the electrical resistivity was decreasing progressively mainly in the near surface (Fig.2C). Resistivity decreased until 80% between the initial time step and the maximum of the rainy event. This strong variation is probably related to the important moistness of the near surface horizons. This is not observed at the deeper part of the sections. Moreover, resistivity seems to increase in several deeper zones. But these increases are calculation artifacts due to the inversion process. They appear commonly in ERT time-lapse sections under electrical conductive zones [7, 8]. In this case, deeper horizons are not truly investigated because the ERT sensibility decreases quickly with depth. This low sensitivity is due to the near surface conductive horizon which concentrates the majority of the electrical current. This incapacity of ERT time-lapse to image the deeper part of the profile is demonstrated in the following part by direct modeling.
Fig. 2: Classical processing of ERT monitoring during the rainy event, Gradient array, 64 electrodes. A: Evolution of mean apparent resistivity during monitoring versus rain. Each brown point represents one ERT section. B: Resistivity model process using DC2DinvRes, color attenuation represents coverage index. B: Percentage change in resistivity from initial model.
3.2 Inversion process limitations There are several possibilities to evaluate the quality of resistivity models obtained by data inversion. Such as the sensitivity and resolution matrix [18], the Deph Of Investigation (DOI) index [23] and the coverage [12]. Some of them where presented and discussed in details by Carrière in [5] for this monitoring campaign. In this paper, we chose to evaluate ERT time-lapse data inversion process limitations by direct modeling (theoretical simulation of the observed measurements). This non automatic solution is very explicit and relevant for a non specialist user.
A resistivity model with near surface conductive horizon was used like reference model (Fig.3A). The resistivity of two blocks in depth (5*5m and 5*3m) is gradually decreased from 20 until 95% of their initial resistivity (Fig.3B) simulating the water presence in depth. For each model, a new apparent resistivity data set was calculated by direct modeling. The same way as for the field measurements, each data set is inverted and variations between this model and reference model (Fig.3A) are calculated (Eq.1) and presented in Fig.3C.
Fig. 3: Sensitivity test by direct modeling using DC2DinvRes, 64 electrodes, Gradient array. A: Resistivity model during rainy event, with near surface conductive zone; B: Same model than “A” with two blocs where resistivity is decrease from 20 to 95%; C: Change in resistivity between reference model “A” and inverted model from “B”.
The results of these tests, presented in fig.3 by direct modeling highlight the incapacity of the inversion process to image deep resistivity variations during a rainy event. It is necessary to reach a deep resistivity variation close to 90%, in order to obtain a measurable resistivity variation at the surface. In other words, ERT surveys have not the sensibility to detect water variation in depth. This low sensitivity is due to the near surface conductive zone which concentrates the majority of the electrical injected current. Then, there are not enough current propagated in depth to characterize the medium. In this way, we can conclude that with actual technology, surface based ERT is not able to detect deep water variation in karst media during a rainy event. If interpretation of monitoring using inversion process is limited, we will see in following part that rough result (apparent resistivity) could provide additional information.
3.3 Change in apparent resistivity During ERT campaigns apparent (measured) resistivity (ρα) analysis is usually neglected. However, these rough results could provide, without calculation artifacts, complementary information to the inverted resistivity. For this monitoring campaign, we analyzed apparent resistivity variations (Δρa) between two consecutive time steps. These variations are normalized by the time (ΔT) between both measurement (ρn and ρn-1) using the following equation (Eq.2).
100 n 1 1 * n
[2]
Thus, the results are presented (Fig.4) in hourly percentage change in ρα with a basic representation of vegetation and soil cover of the profile. Previously recognized karst features [4] are also presented.
Fig.4: A: Hourly change in apparent resistivity between two consecutive time steps. Positioning of fracture detected by GPR and basic representation of vegetation and soil cover. B: Evolution of mean apparent resistivity during monitoring versus rain. Each brown point represents one ERT section.
At the beginning of the rainy event (Fig.4A / section 1), ρα decrease moderately and homogeneously along the section. This resistivity decrease could be related with a moistening of the near surface horizons. After the first rainy event (Fig.4A / section 2), ρα is quickly stabilized. During the following strong rain episode (Fig.4A / sections 4 to 7), moistening process appears very heterogeneous and some zones look like preferential pathways. However, it is impossible to link observed ρα variations in depth with deep moistening process. These observed variations can be directly influenced by near surface variations. Just after the rain (Fig.4A / sections 9 and 10), ρα increases in some zones. That could be related with drainage process. These zones fit with some zones identified before like preferential pathway. This second observation reinforced the hypothesis of preferential pathway in some zones but remains impossible to precise the geometry of these pathways. Other zones where drainage process is not identifiable could be related with zones where soil is thicker and remains moist after the rain. The preferential pathways identified do not seem related with variations of vegetation density or soil cover. However, these zones seem related with fractures previously detected by GPR [4]. It is important to point out that we do not image the pathways but we probably image the effect of pathways above the soil. 4. Discussion and conclusion A unique large scale ERT experiment was carried out during a typical Mediterranean autumn rainy episode (30 days). A total of 230mm of rain was registered and more than 100 ERT time-lapse sections were measured over the same profile during the event. The main goal was to evaluate efficiency and limits of the ERT to monitor water infiltration via previously recognized karst features, under natural conditions. With the first rain, the near surface electrical resistivity decreases strongly due to progressive moistening of soil by the rain. The sensibility of the ERT decreases quickly with depth due to this conductive near surface horizon which concentrates the majority of electrical current. Forward modeling (theoretical simulation of the observed measurements) provided relevant data treatment limitations mainly for the deeper parts of the profile. Consequently, the classic ERT data processing is limited because the inversion process induces important artifacts. Thus, it is necessary to analyze the lateral and vertical variation of the electrical resistivity. Combined observation of apparent and inverted resistivity variations during and after the rainy event highlights some interesting zones. They could be interpreted as preferential pathways, where water dynamic seems quicker in term of moistening and drainage. Nevertheless, these results have to be interpreted reasonably because ERT does not provide enough precision to determine pathways geometry and functioning.
Bibliography [1] Abbas A M, Atya M A, Al-Sayed E A and Kamei H (2004) Assessment of groundwater resources of the Nuweiba area at Sinai Peninsula, Egypt by using geoelectric data corrected for the influence of near surface inhomogeneities. Journal of Applied Geophysics 56(2): 107-122. [2] Al-fares W, Bakalowicz M, Guérin R, and Dukhan M (2002) Analysis of the karst aquifer structure of the Lamalou area (Hérault, France) with ground penetrating radar. Journal of Applied Geophysics 51(2–4): 97-106. [3] Cardarelli E, Di Filippo G and Tuccinardi E (2006) Electrical resistivity tomography to detect buried cavities in Rome: a case study. Near Surface Geophysics 4(6): 387-392. [4] Carrière S D, Chalikakis K, Sénéchal G, Danquigny C and Emblanch C (2013) Combining Electrical Resistivity Tomography and Ground Penetrating Radar to study geological structuring of karst Unsaturated Zone. Journal of Applied Geophysics 94(0): 31-41. [5] Carrière S D (2014) Etude hydrogéophysique de la structure et du fonctionnement de la zone non saturée du karst. Thèse de Doctorat. Université d’Avignon et des Pays de Vaucluse. 210p.
[6] Chalikakis K, Plagnes V, Guerin R, Valois R and Bosch F P (2011) Contribution of geophysical methods to karst-system exploration: an overview. Hydrogeology Journal 19(6): 1169-1180. [7] Clément R, Descloitres M, Günther T, Ribolzi O and Legchenko A (2009) Influence of shallow infiltration on time-lapse ERT: Experience of advanced interpretation. Comptes Rendus Geoscience 341(10–11): 886-898. [8] Clément R, Descloitres M, Günther T, Oxarango L, Morra C, Laurent J P and Gourc J P (2010) Improvement of electrical resistivity tomography for leachate injection monitoring. Waste Management 30(3): 452-464. [9] Dahlin T (2001) The development of DC resistivity imaging techniques. Computers & Geosciences 27(9): 1019-1029. [10] Dahlin T and Zhou B (2004) A numerical comparison of 2D resistivity imaging with 10 electrode arrays. Geophysical Prospecting 52(5): 379-398. [11] Descloitres M, Ribolzi O, Le Troquer Y and Thiébaux J-P (2008). Study of water tension differences in heterogeneous sandy soils using surface ERT. Journal of Applied Geophysics 64(0): 83-98. [12] Günther T (2004) Inversion methods and resolution analysis for the 2D/3D reconstruction of resistivity structure from DC measurements. Freiberg, Germany, University of Mining and Technology. PhD: 160. [13] Günther T, Rücker C and Spitzer K (2006) Three-dimensional modelling and inversion of dc resistivity data incorporating topography - II. Inversion. Geophysical Journal International 166: 506-517. [14] Günther T and Rücker C (2013) Boundless electrical resistivity tomography - BERT 2 - the user tutorial. www.resistivity.net [15] Jardani A, Dupont J P and Revil A (2006) Self-potential signals associated with preferential groundwater flow pathways in sinkholes. Journal of Geophysical Research 111(B9). [16] LaBrecque D J, Miletto M, Daily W, Ramirez A and Owen E (1996) The effects of noise on Occam's inversion of resistivity tomography data. Geophysics 61(2): 538-548. [17] Loke M H (1999) Time-Lapse resistivity imaging inversion. 5th meeting of the environmental and Engineering Society Section. Budapest, Hungary [18] Loke M H (2012) Geoelectrical Imaging 2D&3D. Geotomo software manual. www.geotomosoft.com [19] Loke M H and Barker R D (1996) Rapid least-squares inversion of apparent resistivity pseudosections by a quasi-Newton method. Geophysical Prospecting 44(1): 131-152. [20] Loke M H and Dahlin T (2002) A comparison of the Gauss–Newton and quasi-Newton methods in resistivity imaging inversion. Journal of Applied Geophysics 49(3): 149-162. [21] Masse J-P and Fenerci-Masse M (2011) Drowning discontinuities and stratigraphic correlation in platform carbonates. The late Barremian-early Aptian record of southeast France. Cretaceous Research 32(6): 659-684. [22] Meyerhoff S B, Karaoulis M, Fiebig F, Maxwell R M, Revil A and Martin J B (2012) Visualization of conduit-matrix conductivity differences in a karst aquifer using time-lapse electrical resistivity. Geophysical Research Letters 39(24): L24401. [23] Oldenburg D W and Li Y G (1999) Estimating depth of investigation in dc resistivity and IP surveys. Geophysics 64(2): 403-416. [24] Puig J M (1987) Le système kasrtique de la Fontaine de Vaucluse. Avignon, Université d'Avignon et des Pays de Vaucluse. PhD: 207 [25] Ritz M, Robain H, Pervago E, Albouy Y, Camerlynck C and Descloitres M (1999) Improvement to resistivity pseudosection modelling by removal of near-surface inhomogeneity effects: application to a soil system in south Cameroon. Geophysical Prospecting 47(2): 85-101. [26] Rücker C, Günther T and Spitzer K (2006) Three-dimmensional modelling and inversion of dc resistivity data incorporating topography - I. Modelling. Geophysical Journal International 166: 495-505. [27] Valois R (2011) Caractérisation structurale de morphologies karstiques superficielles et suivi temporel de l'infiltration à l'aide des méthodes électriques et sismiques. Paris, Univ. Paris VI. PhD: 244. [28] Valois R, Bermejo L, Guerin R, Hinguant S, Pigeaud R and Rodet J (2010) Karstic Morphologies Identified with Geophysics around Saulges Caves (Mayenne, France). Archaeological Prospection 17(3): 151-160. [29] Zhou W, Beck B F and Stephenson J B (2000) Reliability of dipole-dipole electrical resistivity tomography for defining depth to bedrock in covered karst terranes. Environmental Geology 39(7): 760-766.
