Sie sind auf Seite 1von 7

Physics and Chemistry of the Earth 36 (2011) 13981404

Contents lists available at ScienceDirect

Physics and Chemistry of the Earth


journal homepage: www.elsevier.com/locate/pce

Which geoelectric array sees the deepest in a noisy environment? Depth of detectability values of multielectrode systems for various two-dimensional models
Sndor Szalai, Attila Novk, Lszl Szarka
Geodetic and Geophysical Research Institute of the Hungarian Academy of Sciences, H-9401 Sopron POB 5, Hungary

a r t i c l e

i n f o

a b s t r a c t
In an urban environment, the two dominating difculties of any geophysical measurement are as follows: (1) the area available for measurements is restricted, therefore, the length of the geoelectric layout is also limited; (2) the noise level is high, and noise diminishes the maximum depth at which any subsurface target can be detected. For this article, the depth of detectability values (i.e. the maximum depth values at which useful information can still be obtained about a given inhomogeneity at a given noise level) were determined from results of two-dimensional numerical model computations assuming multielectrode (electrical resistivity tomography) measurements. Depth of detectability studies were carried out for six different geoelectric arrays, assuming various noise levels. It was found that the depth of detectability is both array and model dependent. Therefore, in a eld study it is advisable to select the array which has the greatest possible depth of detectability values for the given model at the given noise level. When there is no a priori assumption about the subsurface model, either the pole-dipole (P-DP) array or the dipole axial (DP-ax) array is recommended as a default selection, because for most of the studied models they provide the largest depth of detectability values. A twofold increase in the noise level may lead to a decrease of approximately one half in the depth of detectability itself. 2011 Elsevier Ltd. All rights reserved.

Article history: Received 11 March 2010 Received in revised form 11 January 2011 Accepted 19 January 2011 Available online 25 January 2011 Keywords: Geoelectric array Depth of investigation Depth of detection Depth of detectability Multielectrode systems Noise

1. Introduction Depth of investigation is a basic parameter of all geophysical methods, including geoelectrics. The depth of investigation using geoelectric methods was rst attempted to be determined by Evjen (1938) by using the spatial distribution of the currents at depth. Later it became evident that the depth of investigation is inseparable from the selected array. Roy and Apparao (1971) dened depth of investigation characteristics (DIC) of a given array as the depth at which a thin sheet produces a maximum response. Roy (1972) and Bhattacharya and Dutta (1983) extended this approach to further arrays. Alternatively, using the same DIC function, Edwards (1977) recommended the use of the median depth, i.e. the depth at which the integral of the DIC function from the surface to the median depth is the same as for the median depth to innity. Edwards (1977) found this was in better agreement with his eld experience. As was found (Szalai et al., 2009), the median depth can be determined from the RoyApparao depth values by multiplication with 1.59 0.31, which implies a mathematical similarity among various DIC functions. Apparao et al. (1992, 1997) com-

Corresponding author. Tel.: +36 (99) 508 342; fax: +36 (99) 508 355.
E-mail addresses: szarka@ggki.hu, szalai@ggki.hu (L. Szarka). 1474-7065/$ - see front matter 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.pce.2011.01.008

pared so-called depth of detection (DD) values of various inline and broadside geoelectric arrays for resistive and conductive models, by using analogue modelling techniques. In urban geophysics, the depth of investigation is an especially important issue. In cities, the geoelectric measurements should be carried out in a limited measuring area as a consequence of builtup fractions. By decreasing the length of the layout, the depth of investigation will decrease as well. Therefore, it is advisable to nd: (1) those arrays for the multielectrode surveys which have the largest depth of investigation values among the already known arrays (i.e. which are able to see the deepest); (2) new arrays, which have even larger depth of investigation values. The relevancy of the problem is illustrated, among others, by Verma et al. (2009). In this paper, based on earlier work (Novk et al., 2006), an overview is provided about depth of detectability values of several typical geoelectric arrays for basic two-dimensional models at given noise levels. The collection of all published (92 + 10) geoelectric arrays (Szalai and Szarka, 2008a) is suitable for any inter-comparison and for the selection of the most suitable array for any given purpose. A comparison was provided among all known geoelectric arrays on the basis of their parameter sensitivity maps (Szalai and Szarka, 2008b,c) and of their depth of investigation values over thin-sheet models (Szalai et al., 2009).

S. Szalai et al. / Physics and Chemistry of the Earth 36 (2011) 13981404

1399

Normalized depth of investigation (NDIC) characteristics and their resulting depth of investigation values exist only for about 30 arrays of the total number of 92 + 10 (Szalai et al., 2009). It was found that the focussed arrays do not provide as large a depth of investigation as it had been generally thought. At the same time, there are some lesser-known arrays which see deeper than other

arrays. Especially some so-called quasi-null arrays (when the measured potential difference over a homogeneous half-space is nearly zero see the null arrays in Szalai et al., 2002) seem to be promising: those which have with a very at normalized DIC function. Fig. 1 (modied after Szalai et al. (2009); Szalai and Szarka (2009)) illustrates through ve hypothetical cases the general

Fig. 1. DIC functions of various hypothetical arrays. From ae, the DIC function becomes more and more at, which indicates apparently increasing values of the depth of investigation value. In the case of certain null arrays, the DIC function is a constant zero-value function (e). The z/R depth of investigation values belonging to the given DIC functions and the hypothetical noise level are shown as well. Horizontal arrows at the top of the gure point toward increasing values. The maximum values of DIC functions are shown by black dots. Modied after Szalai et al. (2009), Szalai and Szarka (2009).

Fig. 2. A resistivity model and one of its responses, as an illustration of the depth of detectability problem. The white/black dotted line delineates the area in which the relative anomaly is larger than 5%/10%.

1400

S. Szalai et al. / Physics and Chemistry of the Earth 36 (2011) 13981404

behavior of all NDIC functions. It also summarizes, in visual form, the mutual conicts among the RoyApparao depth of investigation values (indicated by black dots), the noise level (shown by dashed lines), and the vertical resolution (the sharpness of the curves). Even the rst two requirements cannot be satised at the same time. As shown in Fig. 1, arrays with NDIC curves shown in Fig. 1c and d see deeper than the arrays having NDIC curves shown in Fig. 1a and b, but there is a large difference between the situations shown in Fig. 1c and d: in case 1c the signal is still over the noise level, therefore this situation is practically useful,

while in case 1d the signal completely melts into the noise. Consequently, an apparently large depth of investigation value, which is obtained from the NDIC function and shown in Fig. 1d, is merely an illusion. In order to be able to select the array having the largest depth of investigation, the level of noise should also be considered. The measured signal should be of higher intensity than the noise. The measurement of the noise level has not been solved yet. Weiss (2009) recently applied electric noise for the characterization of the subsurface in an urban environment; Szalai et al. (2008) investigated the effect of noise originating from positioning

Fig. 3. Geometry of the six studied arrays: pole-dipole (P-DP), dipole axial (DP-ax), Wenner-b (W-b), Wenner-a (W-a), dipole-equatorial (DP-eq), polepole (PP).

Fig. 4. The conductive and resistive variants of the square prism model and the corresponding 100|qextr q1|/q1 values for six DC arrays, as a function of the depth d to the top of the target: (a) conductive anomalous body, (b) resistive anomalous body.

S. Szalai et al. / Physics and Chemistry of the Earth 36 (2011) 13981404

1401

errors (e.g. in rocky surfaces) in multielectrode arrays. The usual noise level in an urban environment is undoubtedly high; the noise level problem should not be neglected. Since the noise level means serious constraints on the multielectrode results, a new denition for the depth of investigation was introduced, in which the noise is also considered. Depth of detectability values (i.e. the maximum depth values at which useful information can still be obtained about a given inhomogeneity at a given noise level) were determined from results of two-dimensional numerical model computations assuming multielectrode (electrical resistivity tomography) measurements. More or less similar investigations had already been carried out by Apparao et al. (1992, 1997). They compared depth of detection values of various inline and broadside geoelectric arrays from prole measurements in resistive and conductive two-dimensional models. These analogue modelling results are compared below with the numerical results obtained by the authors.

must be collinear and evenly spaced, and the resistivity measurements are carried out along a linear measuring prole. The product of a two-dimensional survey is hundreds or thousands of apparent resistivity measurements at various depths beneath the entire length of the electrode array, or what is known as a resistivity pseudo-section. For multielectrode measurements, many different electrode arrays can be used. The features of the obtained resistivity pseudo-section (including the depth of investigation and the noise sensitivity) depend strongly on the applied electrode arrays.

3. Method Assuming various noise levels, a depth of detectability investigation was carried out for six different DC arrays for resistive and conductive variants of three different subsurface inhomogeneity geometries. Due to the fact that nowadays the largest part of DC surveys is carried out by applying multielectrode systems and Lokes RES2DINV software (Loke, 1999), the corresponding forward modelling code was applied: RES2DMOD. The six selected electrode arrays (Wenner-a, Wenner-b, polepole, dipole equatorial, pole-dipole and dipole axial) are among the most frequently used arrays, and at the same time, all of them are included in the RESDINV software. In this paper, the results for 5% and 10% noise are presented. The denition of the depth of detectability is illustrated in Fig. 2. An inhomogeneity at depth produces an apparent resistivity anomaly image. Among the computed resistivity values, the extreme resistivity value qextr was considered: the maximum value for the resistive model and the minimum value for the conductive model,

2. Methodological background An electrode array is a conguration of a few (usually four) electrodes used for measuring electrical resistivity of the subsurface on the basis of Ohms law: simply measuring the voltage between two (potential) electrodes caused by a given current intensity injected into the ground through two other (current) electrodes. In multielectrode systems, such simple measurements are carried out consecutively, and all possible quartets of electrodes can be addressed. In two-dimensional surveys (Dahlin, 1993), the electrodes

Fig. 5. The conductive and resistive variants of the dyke models and the corresponding 100|qextr q1|/q1 values for six DC arrays, as a function of the depth d to the top of the target: (a) conductive anomalous body, (b) resistive anomalous body.

1402

S. Szalai et al. / Physics and Chemistry of the Earth 36 (2011) 13981404

Fig. 6. The conductive and resistive variants of the thin-sheet model and the corresponding 100|qextr q1|/q1 values for six DC arrays, as a function of the depth d to the top of the target: (a) conductive anomalous body, (b) resistive anomalous body.

and 100|qextr q1|/q1 were computed, where q1 is the host resistivity. The model depth varied in the range of 110 m, increasing gradually with a step of 0.5 m. In Fig. 2, the white/black dotted line delineates the area in which the relative anomaly is larger than 5%/10%. Increasing the depth of the model, the area encircled by the black/white dashed lines will be less and less. The depth at which the 10% black dashed lines disappear is called the depth of detectability with 10% noise. At some larger depth, the 5% white dashed line will also disappear; this depth is called the depth of detectability with 5% noise. Below this depth, the inhomogeneity cannot be detected in the case of the given level of noise. The parameters of the forward modelling are as follows: 100 electrodes were applied, and the distance between the neighboring electrodes was 1 m. In the case of Wenner-a (W-a), Wenner-b (Wb), polepole (PP) and dipole equatorial (Dp-eq) arrays, 30 various electrode distances were applied; in the case of pole-dipole (P-DP) and dipole axial (Dp-ax) arrays, the value corresponding to the dipole lengths was a = 1, and 30 different values for the distance between the dipoles were considered. The mesh data are also shown in Fig. 2; the arrays themselves are shown in Fig. 3. In Figs. 46, the 100|qextr q1|/q1 values are shown for resistive and conductive variants of three different models: a rectangular prism (Fig. 4), a dyke (Fig. 5) and a layer (Fig. 6). In each gure, the results for the six aforementioned DC arrays are shown. In these gures, the intersections of the curves with horizontal lines of 10% and 5% directly provide the maximum depth of detectability in the case of the given inhomogeneity.

4. Results In the case of the two narrow models, the P-DP and DP-ax arrays proved to be the best, as summarized in Table 1. The worst results, with one single exception, were obtained by using the PP and W-a arrays, while the W-b and DP-eq arrays were neither the best nor the worst. Between the DD values for different arrays there might be a really large difference, as illustrated by the resistive dyke in Fig. 7: by using the PP array the depth of detectability is 2.1 m, while using the P-DP array, it is 7.1 m. This means that a well-selected array is able to detect a dyke which is at an unreachable large depth for another array. It is easily possible that some unusual arrays out of the existing 92 + 10 arrays would provide even higher DD values than the maximum DD value in Table 1. It can also be seen in Table 1 that by increasing the noise, the corresponding DD values decrease, as expected. The depth of detectability values of various arrays can be compared exclusively for a given model. As shown in Table 1, a conductive prismatic body is e.g. detectable at larger depths than a resistive body. At the same time, in the case of a dyke-like geometry (when the body has a depth extension) and in the case of P-DP and DP-ax arrays, resistive bodies are detectable at larger depths than conductive ones. Meanwhile in the case of both narrow models, the DD values largely depend on the conductive or resistive character of the model: in the case of the thin-sheet model, the resistivity contrast has hardly any inuence on the DD value (Table 1).

S. Szalai et al. / Physics and Chemistry of the Earth 36 (2011) 13981404 Table 1 5% and 10% depth of detectability values of six DC arrays, in meters, for the two (resistive and conductive) variants of the three models, shown in Figs. 46. Array Model Finite width (2D) models Square prism Conductive (q2 = 10 O m) Noise level P-DP DP-ax W-b W-a DP-eq PP 5% 6.6 7.9 6.9 3.6 7.9 5.1 10% 4.2 5.0 4.3 2.0 4.8 3.1 Resistive (q2 = 1000 O m) 5% 5.7 5.4 4.1 2.8 3.0 2.9 10% 3.8 3.6 2.6 1.8 1.9 1.8 Dyke Conductive (q2 = 50 O m) 5% 5.8 5.6 4.9 1.9 4.1 3.6 10% 3.1 2.9 2.1 1.0 1.7 1.5 Resistive (q2 = 200 O m) 5% 7.1 6.9 3.2 2.5 2.1 2.1 10% 4.3 4.0 1.6 1.1 1.1 1.0 Innite width (ID) models ID thin sheet Conductive (q2 = 50 O m) 5% 3.8 4.9 4.2 3.6 3.6 2.5 10% 1.7 2.1 1.9 1.6 1.6 1.0 Resistive (q2 = 200 O m) 5% 3.9 5.1 43 5.1 3.7 2.6 10% 1.8 2.2 2.0 1.7 1.7 1.2

1403

Fig. 7. DD values of a conductive vertical slab (dyke) by using PP and P-DP arrays.

Table 2 Evaluation (order and depth value) of the investigated arrays according to DIC and DD. Array P-DP DP-ax W-b DP-eq PP W-a DIC (R: array length) 3rd (0.160.24R) 4th (0.195R) 6th (0.1R) 2nd (0.25R) 1st (0.35R) 5th (0.11R) DD conductive dyke (5% noise) 1st (5.8 m) 2nd (5.6 m) 3rd (4.9 m) 4th (4.1 m) 5th (3.6 m) 6th (1.9 m)

In the case of laterally extended models, the b-type linear arrays (DP-ax and W-b) are the best. The worst results were obtained again by using the PP and W-a arrays (Table 1). Comparing the DD values with the DIC values taken from Szalai et al. (2009), the two array orders are nearly the opposite, with the exception of the W-a array, which proves to be the worst array in both cases (Table 2). This example emphasizes the difference between an idealized situation (i.e. noise-free measurement and innitely thin layer) and a more realistic case (i.e. noisy measurement and a layer with nite thickness). Many more such investigations would be needed in order to decide whether the quasi-null arrays showing large depth of investigation values in a synthetic model full expectations under realistic conditions. In the case of the square prism models, the P-DP is the best in detection for resistive models, while DP-ax is the most effective in detection for conductive models. Elongating this model downwards leads to the dyke model, where it is again the P-DP array which sees the deepest, regardless of the resistivity contrast. Elongating the square model in the horizontal direction leads to the thin-sheet model, where the DP-ax array is the best, regardless again of the resistivity contrast (Table 1). It is remarkable that the six 100|qextr q1|/q1 curves belonging to the six arrays, as a function of the model depth, never (i.e. in the case of either model) cross each other. It means that the order of the arrays is independent of the actual model depth. The difference among the six curves is the smallest in the case of the thin-sheet

model, and it is the largest in the case of the dyke model. Here, and also in the case of the square model there are two better and four worse arrays. The two outstanding arrays are in any case the P-DP and the DP-ax arrays. Comparing these results with those by Apparao et al. (1992, 1997), who used a slightly different denition for depth of detectability (and called it depth of detection), the only discrepancy is obtained with the PP array. In this study, the PP array, in terms of the depth of detectability value, does not hold such a distinguished place as had been found by Apparao et al. (1992, 1997). As clearly seen in Fig. 4 by Apparao et al. (1997), their physical modelling, evidently in a nite-size tank, did not make it possible to cover the whole range of the geometry. In the optimum data set obtained with four-electrode arrays by Stummer et al. (2004), in the rst 600 places there are non-standard dipoledipole-type arrays. The DP-ax and Wenner-b arrays are also dipoledipole-type arrays, which proved to be signicantly better than the other two investigated four-electrode arrays (the DP-eq and the Wenner-a). It should be noted that incorporation of some three-electrode arrays into multielectrode systems would certainly improve the multielectrode interpretation results, especially at the deepest parts of the pseudo-section. According to model and eld studies by Candansayar, 2008, a joint inversion of PD-L (namely P-DP), a PD-R (namely DP-P) and DP-ax arrays provides a more realistic underground resistivity structure than an inversion of any three- or four-electrode arrays. The results herein implicitly support the conclusions of Candansayar, 2008.

5. Conclusions In this paper, an overview is provided about depth of detectability values of various geoelectric arrays for the basic two-dimensional models at given noise levels. The depth of detectability values depend strongly on the model parameters themselves. In the case of inhomogeneities having small lateral extension, the P-DP and DP-ax arrays proved to be the best. The worst results, with one single exception, were obtained by using the PP and the W-a arrays. In the case of these narrow inhomogeneities, the W-b and DP-eq arrays proved to be neither the best nor the worst arrays. In the case of the laterally extended models, the b-type four-electrode arrays (DP-ax and W-b) are the best. The worst results were obtained again by using the PP and W-a arrays. The order of arrays based on DD values is signicantly different from the noise-free DIC-based order: the arrays are practically, with the exception of the W-a array (which is the worst array in both lists), in an inverse order.

1404

S. Szalai et al. / Physics and Chemistry of the Earth 36 (2011) 13981404 Lund Institute of Technology, Lund University, Sweden, pp. 187. ISRN: LUTVDG/ TVTG1007SE. SBN 9162810324. Edwards, L.S., 1977. A modied pseudosection for resistivity and IP. Geophysics 42, 10201036. Evjen, H.M., 1938. Depth factors and resolving power of electrical measurements. Geophysics 3, 7895. Loke, M.H., 1999. RES2DMOD ver. 2.2. Rapid 2D resistivity forward modelling using the nite difference and nite-element methods. Wenner (alpha, beta, gamma), inline & equatorial dipoledipole, polepole, pole-dipole and WennerSchlumberger. Freeware courtesy of M.H. Loke. <http://www.abem.se/les/ res/manr2dmod.pdf>. Novk, A., Szalai, S., Szarka, L., 2006. Target depth of detectability of DC arrays for various models. In: 12th European Meeting of Environmental and Engineering Geophysics, Helsinki. <http://www.earthdoc.org/detail.php?pubid=637>. Roy, A., Apparao, A., 1971. Depth of investigation in direct current methods. Geophysics 36, 943959. Roy, A., 1972. Depth of investigation in Wenner, three electrode and dipoledipole DC resistivity methods. Geophys. Prosp. 20, 329340. Stummer, P., Maurer, H., Green, A.G., 2004. Experimental design: electrical resistivity data sets that provide optimum subsurface information. Geophysics 69 (1), 120139, 10.1190/1.1649381. Szalai, S., Szarka, L., 2008a. On the classication of surface geoelectric arrays. Geophys. Prosp. 56, 159175. doi:10.1111/j.1365-2478.2007.00673.x. Szalai, S., Szarka, L., 2008b. Parameter sensitivity maps of surface geoelectric arrays, I. Linear arrays. Acta Geod. Geoph. Hung. 43 (4), 419437. doi:10.1556/ AGeod.43.2008.4.4. Szalai, S., Szarka, L., 2008c. Parameter sensitivity maps of surface geoelectric arrays, II. Nonlinear and focussed arrays. Acta Geod. Geoph. Hung. 43 (4), 439447. doi:10.1556/AGeod.43.2008.4.5. Szalai, S., Szarka, L., 2009. Geoelectric arrays. Curr. Sci. 96 (10), 1298. Szalai, S., Szarka, L., Prcser, E., Bosch, F., Mller, I., Turberg, P., 2002. Geoelectric mapping of near-surface karstic fractures by using null-arrays. Geophysics 67, 17691778. Szalai, S., Koppn, A., Szarka, L., 2008. Effect of positional inaccuracies on multielectrode results. Acta. Geod. Geoph Hung. 43 (1), 3342. Szalai, S., Novk, A., Szarka, L., 2009. Depth of investigation and vertical resolution of surface geoelectric arrays. J. Eng. Environ. Geophys. 14, 1523. Verma, S.K., Manglik, A., Krishna Murthy, N., Ananda Rao, V., Bhatt, K.M., Chandra, S., Tezkan, B., Harinarayana, T., Scholl, C., Patro, P.K., Dutta, S., 2009. Comparative efciency of various electrical and electromagnetic methods in mapping shallow 3-D conductors encountered in urban geophysical problems. Eos Trans. AGU 90 (52), Fall Meet. Suppl., Abstract NS22A-04. Weiss, C., 2009. More noise, please: how cultural overprinting in the urban environment can be exploited for improved subsurface imaging (Invited). Eos Trans. AGU 90 (52), Fall Meet. Suppl., Abstract NS22A-01.

The DD values of various geoelectric arrays for a given model cover a wide range: there can even be a difference of 34 times between the maximum and minimum values. Therefore, it is very important to apply the most favorable geoelectric array in an urban environment. Due to the model dependence of DD values, if there is some a priori information about the subsurface model, it is recommended to select the geoelectric array for the eld measurements from Table 1. Otherwise, in general, it is recommended to select either the P-DP or the DP-ax arrays. The numerical study by the authors has made it possible to select the arrays having the largest possible depth of detection both in the case of and in the absence of a priori information about the investigated subsurface target. This is especially important in an urban environment, where the measuring length is often limited and the noise level is high. Acknowledgments Hungarian National Research Fund, Project numbers K49604 and NI 60013. One of the authors of this paper, S. Szalai, is a grantee of the Bolyai Jnos Scholarship. References
Apparao, A., Gangadhara Rao, T., Sivarama Sastry, R., Subrahmanya Sarma, V., 1992. Depth of detection of buried conductive targets with different electrode arrays in resistivity prospecting. Geophysical Prospecting 40 (7), 749760. Apparao, A., Sivarama Sastry, R., Subrahmanya Sarma, V., 1997. Depth of detection of buried resistive targets with some electrode arrays in electrical prospecting. Geophysical Prospecting 45 (3), 365375. Bhattacharya, B.B., Dutta, I., 1983. Depth of investigation studies for gradient arrays over homogeneous isotropic half space. Geophysics 48, 12481251. Candansayar, M.E., 2008. Two-dimensional individual and joint inversion of threeand four-electrode array DC resistivity data. J. Geophys. Eng. 5, 290300. doi:10.1088/1742-2132/5/3/005. Dahlin, T., 1993. On the automation of 2D resistivity surveying for engineering and environmental applications. PhD thesis. Department of Engineering Geology,

Das könnte Ihnen auch gefallen