IFEI IEU F_,-IEI -IVSIE5 ELSEVIER Journal of Applied Geophysics 33 (1995) 77-91 Analysis of GPR data: wave propagation velocity determination Sylvie Tillard a, Jean-Claude Dubois b a Direction du Cycle du Combustible, D~partement d'Entreposage et de Stockage des Dgchets (CEA/DCC/DESD), 13108 Saint Paul lez Durance cedex, France b lnstitut Frangais du P~trole, BP 311, 92506 Rueil Malmaison, France Received 27 November 1992; accepted 27 December 1993 Abstract The detection of millimeter-wide discontinuities such as fissures in subsurface geological formations may be possible by means of GPR soundings, but establishing a law for the electromagnetic wave propagation velocity from field data in order to interpret radar data and to localize these discontinuities in depth is not easy. In order to optimize the interpretation of such radar surveys carried out at various sites, we turned our attention to the accuracy by which the electromagnetic wave propagation velocity may be determined. In a granitic quarry, with the help of a borehole cored 40 m deep, we were able to appraise the limits of velocity analysis based on normal move-out corrections. We characterized the degree of error caused by slightly dipping reflectors and we showed the instability of interval velocity calculations caused by uncertainties in reflection time and RMS velocity assessment. The velocity profile derived from laboratory measurements of dielectric permittivity of the granite samples, for various depths of cores, proved to be insufficient for an exact mapping of the fissure network. At a limestone quarry site, we showed that by working on several paths of electromagnetic waves, direct velocity determinations could help in differentiating rocks in the same formation, based on quality of the field data. Finally, velocities calculated in an anisotropic environment and in geological formations covered by a road or a concrete surface were analysed. Using data recorded in schists, we obtained an example of a non-uniform distribution of velocity in the same formation. Using data recorded in sands and in granite, we demonstrated that a surface material had negligible effect on the determination of the subjacent geological environment velocity. I. Introduction Ground penetrating radar is becoming increasingly successful in geophysics and in civil engineering. Many effective radar surveys and results are being presented in highly diverse fields o f application. However, many users interpret the records using only a superficial examination of the images to simply specify that an object has been detected or that a reflector exists. A depth scale is sometimes indicated but depth uncertainty margins are seldom added. The wave propagation velocity may have been roughly deduced from dielectric permittivities published in the literature without taking into account the complex nature of this 0926-9851/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved S S D I 0 9 2 6 - 9 8 5 1 (94)00023-H parameter in a lossy environment. When the measuring system allows the transmitting antenna to be physically isolated from the receiving antenna, wide angle or common mid-point measurements may be carried out to determine the velocity on the basis of assumed direct path of the underground wave and then applied to the whole formation. In the case of a fairly homogeneous environment such as a granitic block or a salt formation, this approximation is generally realistic but it provides only a rough estimate for multilayer terrains having a variable porosity or variations in water saturation. To assess the quality of the depth scales usually indicated, we examined various possibilities o f determining radar velocities. Our aim was to look for any 78 S. Tillard, J.-C. Dubois / Journal (~[Applied Geophysics 33 (1995) 77 91 difficulties inherent in the methods and to define their limitations. The importance of this survey subject should not be underestimated since, in the case of positive conclusions about target localization accuracy based upon improved velocity data, GPR could be considered as a continuous non-destructive prospecting method, theoretically allowing destructive soundings and soil samplings to be suppressed (or at least decreased) for calibrating the surveys with the geological characteristics. In addition, we will also consider the type of information that can be extracted from the velocity values in order to help characterize the propagation environment. In this paper we present measurements carried out in the laboratory to determine the real and imaginary components of dielectric permittivity of cored rock samples and we compare these results with radar field data. With regard to the latter, processing methods borrowed from seismic data analysis were applied because of the direct similarity between electromagnetic and acoustic propagation (Tillard, 1991; Dubois, 1992; Fisher et al., 1992). The advantage of basic processing consisting, for instance, in spectral analysis and various filtering is fairly obvious. On the other hand, the use and merits of velocity analysis remains to be demonstrated with respect to GPR data. Most of the data presented in this work were collected with a PC-based, digital radar system ~, in a granitic quarry where we had a 40 m deep cored borehole at our disposal in order to validate the techniques used to calculate velocity vs. depth. Other data from experiments performed on limestone, glacial, sandy, and schistous formations are also studied. With the radar system used in the measurements, three frequency ranges centered on 50, 100, and 200 MHz were available. The transmitting antenna was independent from the receiving antenna in all of the tests. 2. D e t e r m i n a t i o n of a velocity vs. depth law 2. I. Veloci O, analysis in radar sounding, two kinds of velocity measurements can be carried out depending on whether the Pulse E K K O IV, manufactured by Sensors and Software, Mississauga. Canada. antenna offset is fixed or can be raised. Wide angle ( W A ) reflection or common mid-point (CMP) measurements provide the data necessary for calculating propagation velocity. In the first case, one antenna remains stationary while the other is moved along the profile direction. In the second case, both antennas are simultaneously moved apart on either side of the midpoint of the profile. Below, a review of the equations governing the wave path for these radar sections are presented. In the case of a horizontal reflecting plane in a homogeneous medium, the two-way bistatic traveltime of the reflected wave can be written as: x ~ 4h ~ t2 =--4 + t, ~ (1 l, 2 where t is two-way traveltime ( s ) ; x is offset (m) ; h is mirror depth (m) ; and r is propagation velocity ( m / s). For each hyperbola of the form indicated by Eq. 1 ), a traveltime, A to, is defined by the difference between the two-way traveltime for a given offset t(x) and the vertical two-way traveltime at the origin to for x = 0 (normal move-out [NMO] corrections). At~=t(x)-to=to +~- 1 (2) where At~ is the NMO correction (s). From this relation, when the offsets are small in comparison with the depth of the refector: X2 At~,~ (3) 2r2to In the case of an inclined reflector, if there is no dip correction, the hyperbola slope overestimates the propagation velocity. In this case, the corrected two-way traveltime is: t- = ~ ( x - + 4h- + 4hx sing0 (4) l!- and I/2 (~'/cosq~)-] (5) where q~is the dip of the reflector (rad). In practice, vertical variations of velocity are generally observed. It can be shown that, for an environ- s. Tillard, J.-C. Dubois / Journal of Applied Geophysics 33 (1995) 77-91 ment made up of several horizontal homogeneous layers and when the transmitter/receiver offset distance is small compared with the investigation depth, the velocity value in Eq. (1), referring to path conditions which are unknown to us, can be approximated by a root mean square (RMS) velocity given by: N (6) VZN= ~_, t, v2~/t where vN is the root mean square velocity between the surface and the base of layer N, ( m / s ) ; t is the vertical one-way traveltime from the surface to the base of layer N (s); tn = vertical one-way traveltime in layer n (s); and vn wave propagation velocity in layer n ( m / s ) . RMS velocities are frequently defined in seismic studies by means of a process called "velocity analysis". This process consists of a computerized move out correction Atc for a succession of velocity values starting from the slowest up to the fastest likely to be encountered in the prospected field. By this' 'hyperbola flattening" technique, we numerically specify the hyperbola along which the signals corresponding to a reflection on a particular reflection point are placed. In other words, in the processed radar section scans, the correction, Ate, corresponding to the velocity at which the reflector appears most accurately horizontal, is sought. If the velocity is too high, the reflector dips downward; if it is too low, it dips upward. Having determined the RMS velocities for various vertical two-way traveltimes, we can use Dix's formula (Yilmaz, 1988) to calculate the velocity in a given layer: t 2 - V~-- NVN--~-I ts-ts-i ~2 N--I (7) where VNis the interval velocity ( m / s ) ; vN is the RMS velocity up to the bottom of layer N ( m / s ) ; vN-l is the RMS velocity up to the bottom of layer N - 1 ( m / s) ; tN is the vertical two-way traveltime up to the bottom of layer N (s); and tN_ t is the vertical two-way traveltime up to the bottom of layer N - 1 (s). For parallel dipping reflectors, the following approximation in Dix's formula is adopted: 2 Vu- t my 2m COS2 q~-- tN_ 2 i/~N-- I tN--tu-i COS2~o (8) 79 The relations concerning the wave path presented above involve terrains with elementary planar geometry. Although complex reflection models are generally necessary in order to develop a practical velocity vs. depth law, we start by testing the simple cases referred to here, using data from our radar records. In some situations, it must be pointed out that we often must apply the NMO corrections under marginal conditions because of the low ratio of investigation depths to the chosen offsets. Indeed, if we take the example of a granitic environment, our GPR system allows measurements to a 25-30 m depth, whereas the antenna offset distance may be up to 20 m so as to ensure a reliable calculation of the direct underground wave velocity by a simple assessment of the time~listance curve slope. The data processing programmes used are borrowed from GEOMAX or GEOVECTEUR software 2. The programmes imply working with distance and time quantities expressed in meters and milliseconds. Rather than resample the radar traces, the input parameters are deliberately distorted. The fictitious correspondence is as follows: a time sample on a radar record is equal to 1 ms. This detail is noted here only to justify the nonrounded values of frequencies and velocities mentioned later in the paper. 2.2. Example o f application Our experimental granitic site presents nearly all the ideal geological conditions required for the tests referred to above, since the rock was stratiform: the quarry was made of a succession of " b e d s " we shall call, as a first estimate, subhorizontal. In the constant offset radar records shown in Fig. 1, in which the horizontal and vertical scales are highly disproportionate, a variable inclination is discernible but it is not very marked (determined to be 5-15 ° of apparent dip; maximum 20°). The bed thickness increases with depth (some decimeters at the surface, 2-3 m at - 4 0 m). Fig. 2 shows a block-diagram of the granite quarry formation. The beds are limited by joints or fissures of just a few millimeters. It was possible to locate these in depth by means of the 40 m cored borehole. Some of the fissures were very apparent in the radar sections 2CGG registered trade mark; software available at the French Petroleum Institute,Rueil Malmaison,France. 80 S. Tillard, J.-C. Dubois / Journal of Applied Geophysics 33 (1995) 77 ~)1 Distance fro) 10 20 30 40 10 20 10 16o ~3 E 320 480 Q ® ® ® 26 m ® (9 4 ®,h," o 46m ! | o : cored borehole ® Fig. 1. Constantoffset 100 MHz radar data, recorded in three directionsin a granitequarry. recorded for all of the offsets and they constitute the reflectors which are of interest to us in this survey. At this granitic site, the piezometric level was 8 m deep. Fig. 3a shows an example of raw data collected in the field to determine radar propagation velocity. Interfering reflections which preceded the direct under- S. Tillard, J.-C. Dubois / Journal of Applied Geophysics 33 (1995) 77-91 2 T',o ~ J 15- ~to Profil 3 Fig. 2. Schematic illustration of the geology of the stratiform granite experiment site: only the main subhorizontal beds limited by joints or fissures are plotted. ground waves were eliminated by operating a " m u t e " before applying NMO corrections. These reflections could have been processed but they would make the operation unnecessarily more complex. In Fig. 3b, an extract of the NMO corrections is presented: only data corrected according to the procedure described above with velocities for the interval 75-162.5 m//xs are 81 shown ( 12.5 m//xs step, eight sections). In this figure, it can be seen that all the reflection hyperbola picked up between 80 and 400 ns are not flattened for a unique velocity value. For the data set in Fig. 3a, the most interesting sections correspond to the 112.5-125 m//xs corrections in Fig. 3b. For the complete process, we extended the velocity range from 50 to 300 m//xs; 300 m//zs is not a "geological" velocity, it corresponds to radar propagation in the air. This value was nevertheless tested since, in some environments, airborne reflections do occur. The latter are recorded because the dipole antennas of the EKKO system are not shielded. The technique described here allows airborne arrivals to be identified since only these reflections will be correctly flattened for an NMO correction corresponding to 300 m//xs. If their arrival times are known, they can then be localized on constant offset records and therefore they will not be mistaken for geological events. As explained above, velocities "calculated" at different depths (or times) lead to a RMS velocity profile corresponding to a single field acquisition (one frequency, one profile direction). We repeated the operation for the fourteen WA or CMP sections available at the granite site for the three prospection frequencies and the three selected profile directions. We thus obtained fourteen RMS profiles; we then superposed these profiles to define the RMS velocity law vs. depth to be used at our experimental site to calculate interval velocities. At first sight, the superposition was very disappointing. For the 50-350 ns time interval, a very large discrepancy was noted for velocities in the 100 and 130 m//xs range, and even more exceptionally up to 140 to 150 m//xs. An average law could therefore not be derived because the uncertainty was too large. Indeed, in the cores of the borehole drilled in the quarry where the fissure density was about 1.3 fissures/m, locating and differentiating fissures which reflected the electromagnetic signal from those which did not on radar sections could not be achieved if we work with an uncertainty about + 15 m//xs. This is illustrated by the following numerical example: if we take a 230 ns twoway traveltime, a 120+ 15 m//xs velocity leads to a depth between 12.1 and 15.5 m. In order to correctly calculate interval velocities, analyses must be gathered according to the elementary parameters we neglected 82 S. Tillard, J.-C. Dubois / Journal of Applied Geophysic.s 33 (1995) 77 9 I Normal Move Out correction 0 0 I 75 87.5 ' i 0 100 (m/p.s): 112.5 +! 125 137.5 150 162.5 ri ;' Distance (m) c! ~9 [... 160 16(] .,.":. ~ ..,-.4-~ ~ 820 320 480 480 (a) (b) Fig. 3. (a) CMP 100 MHz radar data recorded in the granite quarry formation. (b) Extract of NMO corrections performed on the data from (a). above. We shall begin by studying the frequency contribution. A priori, no frequency behaviour can be noted when we compare Fig. 4a, b, and c in which RMS profiles obtained through NMO corrections are classified according to the prospection nominal frequencies. The relative stability of velocity with frequency is confirmed, at least for the direct propagation through the rock, by calculations of the time-distance slopes read on sections provided by the spectral analysis of a CMP data set. Table I gives these slopes for various dominant frequencies, indicating that there is no regular dependance of velocity on frequency. However, a very weak dispersion could occur at the lower end of the frequency range ( 30-60 MHz) because of the slight decrease in dielectric permittivity we derived from laboratory tests on vertical core samples. Fig. 5 presents the results of measurements on dry and water-saturated granite samples. The fact that this dispersion is not discernible in terms of velocity is justified by referring to the uncertainty range affecting the velocity calculations based on the slope determination of time-distance curves. Twenty-eight laboratory measurements carried out using a coaxial transmission line sample holder designed to contain centimeter-sized rock samples give fairly uniform average values of dielectric permittivity ranging from 6.6 at 30 MHz to 6.0 at 60 MHz (Tillard, 1991). These figures indicate vertical propagation velocities ranging from 117 to 122 m / ~ s , but the corresponding values derived from the radar sounding sections overlap because of our error bounds. Depending on the quality of our raw data, the velocity uncertainties fluctuate from _+2 to _+5 m/~s. The previously noted velocity profile disparity is reduced when we separate the results, profile by profile, by WA measurements and by CMP measurements as shown in Fig. 6. in contrast to our first assumptions, velocities vary differently according to the line of measurement and we observed that the W A acquisitions led to RMS velocities that were systematically lower than the CMP acquisitions. Because of the different operating principles for the W A and CMP analysis, in the case of an inclined reflector, the distance h which appears in the correction formula, i.e. the perpendicular S. Tillard, J.-C. Dubois / Journal of Applied Geophysics 33 (1995) 77-9l 83 (a) 1601 140[ 12o t%" ..¢~ " -- ,=--~ - 2 - -6~ ,~--.b:::-::.:'--"Q--:,i-- - -:_-__--t > 100 s'o lbo 1so 260 2so 36o 3so 400 Two-way travel time (ns) W ~160 I (b) 140 I I loo I 5'0 1,50 160 2()0 250 300 350 400 Two-way travel time (ns) • 1601 . 1401 '~' 1201 • (c) ..A ...- .--~ :......D......~o 1oo .........~ 1 'o ~ 260 2s0 360 3so Two-way travel time (ns) Fig. 4. RMS velocities calculated for various directions in the granite quarry, plotted vs. the prospection frequency; (a) 50 MHz; (b) 100 MHz; (c) 200 MHz. to the reflector occuring at an offset mid-point, is not a constant parameter for the WA measurements. Although we thought it was possible to disregard this difference, we now conclude that even slight dips have a disturbing influence. We shall therefore work with CMP data for which we know that the effect of slight dips is minimised and partially rectifiable according to the relation given by Eq. (8). When we consider uncertainty calculations obtained using Dix's formula, we can expect a substantial instability of interval velocities both in seismic (Yilmaz, 1988) and radar surveys. Our experiments are no exception. The sources of error to be taken into account in the velocity analyses are: ( 1 ) determining the origin time to in Eq. (3), which may not correspond to the exact top of the hyperbola; (2) appraising the horizontality of a corrected reflector: this operation can be in error because of the large discrete 12.5 m//zs scan step chosen for the convenience in the NMO hyperbola corrections; and (3) working with large offsets which favour the existence of reflected-refracted waves or lateral waves at ground surface (Clough, 1976): they are all parallel to the air wave but with different delay time and they interfere with the hyperbola slopes. We do not control this third source of error which seems difficult to detect. All of these factors are far from being insignificant. This is illustrated by Fig. 7. We took into account the 15° dip and we converted two similar RMS velocity profiles, shown in Fig. 7a, into interval velocities. The poor quality of our conversion can be noted in Fig. 7b. We build Fig. 7c from Fig. 7a data by grouping the terrain layers that provided the most discrepant results in Fig. 7b. We only considered one layer in the 100200 ns range. We justify this operation by specifying that if a large uncertainty exists for locating the depth in time of a layer on a radar record and if it provides significant interval velocity variations because of the "instability" of Dix's formula mentioned earlier, this layer probably reflects only very little electomagnetic 84 S. Tillard, J.-C. Dubois / Journal of Applied Geophysics 33 (1995) 77-91 Table I Direct underground wave propagation velocity calculated from spectral analyses of Wide Angle data ( granitic quarry ). Frequency (MHz) Velocity calculated on rawdata (m/#s) Filter applied on raw data: ~],f2,f~,f4)" Velocity calculated on filtered data (m/p~s) (MHz) 5(1 117±2 100 117 +2 200 123 _+5 (11,14,17,20) ( !6, 21, 25, 30) (24, 31, 37, 45) (36,46,57,67) (56, 69, 87, 100) (84, 105, 134, 154) ( 13t, 156, 194, 219) (27, 35, 42, 50) (42, 51, 65, 74) (64, 76, 99, 111) (97, 112, 126, 141) ( 125, 142, 179, 192) ( 57, 67, 89, 99) (87, 100, 119, 131) (117,132,164,179 ~ (162, 180, 225,242) ( 224, 244, 302, 322 ) ( 300, 325,400, 425 ) ' " I It) +3 118±2 12t ±2 1i7t4 118±5 120±3 119.+_2 1233-2 124±3 II9±4 '~ 121 ~5 119±5 124±5 121 ± 5 I 19 ± 5 '~= frequency. energy. Therefore, we conclude that it constitutes a minor velocity discontinuity. As a result, a simple law 7i ® ® "o v 20 1oo 5o0 Frequency (MHz) Fig. 5. Dielectric permittivity measured on a granitic sample: (a) in the dry state; (b) saturated with demineralized water. describing velocity vs. depth is indicated. It is plotted in Fig. 8 (top, continuous line) by taking into account the remarks generally made in seismics: interval velocities must be used without neglecting the trends read on R M S profiles. In Fig. 8 ( t o p ) , we compare the velocity law mentioned above with the velocities calculated, on the one hand, using the reflector depths read (in time) at the hole abscissa for constant offset radar sections and, on the other hand, using the bed thicknesses, measured on the cores. Fig. 8 ( b o t t o m ) shows, in addition, the laboratory measurements carried out on saturated rock samples, taken from the peripheral zones of the fissures; the permittivity values have been converted into propagation velocity. W e can draw two conclusions from these comparisons: (1) Although the permittivity measurements are reliable for the characterization of different rock samples since we observe, for instance, a C R I M 3 behaviour for dielectric permittivity vs. porosity (Shen, 1985; Coutanceau, 1989; Tillard, 1991 ), they are not sufficient to account for global variations of velocity vs. ~Complex Refractive Index Method. S. Tillard, J.-C. Dubois / Journal of Applied Geophysics 33 (1995) 77-91 85 Profile 160 ~ 140 .Z' "~ 120 m tlXl fie 800 L I 50 1O0 I L I 150 200 250 Two-way travel time (ns) I 300 3so 400 Profile 2 160 E .Z, •~ _o 140 120 > O9 100 n'- 8O 0 I 0 I 1 O0 I I I 150 200 250 Two-way travel time (ns) I 300 I 350 400 Fig. 6. RMS velocities calculated for the granite quarry formation on wide angle (11) and common mid-point (O) data for two lines of measurements. depth in an environment such as the stratiform granite at the quarry site. The electrical discontinuities are due to fissures themselves; the contribution of the fissure peripheral zones is insignificant. The use of this type of measurement would be more advantageous in the case of propagation through two distinct materials. (2) The velocity law established from the cored borehole presents large variations shown in Fig. 8 (top). Even if we take into account the fact that variations can be affected by an error in reflector depth read on the constant offset radar records, and that this error might be dependent upon the reflection intensity and upon the density of the surrounding fissures, our velocity analysis derived from RMS velocities clearly does not allow so many irregularities to occur over such short vertical distances. Even strong contrasts can be overlooked: this is particularly the case of the decreasing velocity (35 m//zs) observed at 50 ns which can account for the existence of a slow guided dispersive wave appearing on some of our CMP or W A R R sections. 2.3. Discussion We made a point of repeating field measurements to make our survey as complete as possible so as to be able to come to a definite conclusion on velocity analysis. It is important to note that we worked with records showing several distinct reflection hyperbola which we sought to very accurately situate in time. Despite these favourable assumptions, the resulting velocity analyses are only partially utilizable. Without a borehole to derive velocity vs. depth directly, extrapolating the observed surface velocity to the whole massif would have led, for a 200 ns one-way traveltime for instance, to a 4 m underestimation of the depth. Our velocity analysis reduced this apparent depth error to 2 m, allowing some improved data migration tests as shown in S. Tillard, J.-C. Dubois /Journal o[Applied Geophysics 33 (1995) 77-91 g6 (a) loo I o if) rr 1% s'o 16o 1,~o 26o Two-waytraveltime (ns) 25o 360 350 (b) 16o~ 140 I loo r $ 5'o 1so 260 Two-waytravel time (ns) 1~ e- ~ 250 360 350 (c) 16o t .~140 I • ..... • | A. . . . . . . . . . . . . . . o 120 I loo 5'0 160 1,50 260 2Ei0 Two-way travel time (ns) ~0 350 Fig. 7. (a) RMS velocities calculated/'or the granite quarry fonr~ation. (b) Interval velocities derived from the RMS velocity profile. ( c ) Interval velocities derived using two time intervals combined. °1i Ol ~o 360 ~o Boo Two-way travel time (ns) A 135 .~, 125 o o 120 11.~ • AA && s • 1'0 • • ~'s • A ~ ~ 30 35 Depth (m) Fig. 8. Variations of velocity vs. depth in granite quarry formation, O - O = velocities calculated with time depths read on records and with thicknesses of granitic banks measured on cores; - - = velocity vs. depth law, deduced from interval velocities (cf. Fig. 7c); * = velocity calculated from laboratory dielectric permittivity values vs. depth. S. Tillard, J.-C. Dubois/ Journal of Applied Geophysics 33 (1995) 77-91 87 Distance (m) I0 o ,, • ;i 20 I ,i' ~! 40 30 I [ • i ~ I! ,I ,!' L t: ,I "~ q ~r~3 480 Distance (m) 0 g 1Q 20 40 30 LLt ,IILLL LI Lt,LLIU!JJ,I:I:J iI 0 : v.i? , , : . , , : i~ ' i!f , , J- Fig. 9. Migration test ( 100 MHz data) for granite quarry formation.Top: unprocessedrecord; bottom: migrated record. Fig. 9. Mapping multiple fissures cannot be envisioned with any accuracy if it is based upon these velocity data. As a guide, in our case an assessment of the fissure density gave 1.3 fissures/m in the 0-30 m range. The type of processing we studied here is not wholly satisfactory, but we do know now that, if it is used, working with a velocity profile similar to that deduced from the RMS values is enough to establish a representative velocity vs. depth profile providing that the antenna offsets are small in comparison with the reflector depth. In contrast, because of the strong instability of interval velocity calculation caused by uncertainties in the reflection time and RMS velocity assessment, velocity analysis using Dix's formula is not an advantageous process. 3. Utilization of direct traveltime arrivals 3.1. Rock differentiation in a limestone formation Conclusions identical to those given above concerning the efficiency of NMO corrections were deduced from the use of CMP measurements recorded on two levels of a limestone quarry, cut into descending benches. At this site, the limestones are in the form of horizontal beds, a few meters thick, separated by joints of a few centimeters thick, generally filled in by decalcification clays; corresponding altered or fissured zones can measure up to 1 m thick. Here we shall not be concerned with the accuracy of determining the depth or thickness of the multiple beds but rather with the possibility of differentiation by GPR S. Tillard, J.-C. Dubois /Journal of Applied Geophysics 33 (1995) 77-91 88 Distance (m) o Distance (m) o "i. 0 to , 10 .•z - . . . ,,,,,It--... - 0 s 10 20 "'~Ii 200 2oo i.- " E .~ 20 o , 200 Distance (m) . ? |lill|ltl .... 40O 400 400 :!(a) Co) (c) Fig. 10. Velocity measurements for three geological formations with different electrical properties: ( a ) sand and clay; t b ) limestone; ( c ) ice. of two kinds of limestone, one oolitic and one comblanchien outcropping, respectively, at benches 1 and 2. Because of the uncertainties previously mentioned concerning NMO corrections, RMS velocities wilt not be derived since the variations we are looking for are, a priori, slight if we consider the results provided by electric and acoustic loggings ( Dubois, 1992). Instead, we focus attention on the information to be obtained from the utilization of the direct wave path. The CMP profiles recorded on benches 1 and 2 of the quarry give surface velocities of 77 + 4 and 85 __+4 m//xs, respectively. According to measurements of the rock dielectric permittivity, if the difference of 8 m//zs between these layers is real and not attributable to measurement uncertainties, there is an apparently larger porosity for the comblanchien limestone than for the oolitic limestone. It therefore seems possible to apply this test to limestone differentiation. To check this point, we used a rough simulation of a vertical seismic profile performed on the wall separating the two benches. For this purpose, one radar antenna remained stationary on the top level, 2.5 m from the edge of the wall, whereas the second antenna was moved down the wall at 0.5 m intervals. The use of the oblique wave path on the records thus obtained yielded an average velocity of 80 m//xs, a velocity value compatible with the previously given values. Indeed, if we attributed a propagation velocity of 85 m / ~ s to comblanchien limestone and 77 m//,~s to oolitic limestone, the path of the radar waves through 4 and 8.5 m of those different limestones, respectively, results in a two-way traveltime of 315 ns. In the earlier tests we also found this time to correspond to an average vertical velocity of 80 m//~s for a total path of 12.5 m. By means of these tests, we showed that a study of direct propagation velocity would allow the differentiation, in a same formation, of rocks of similar quality on the basis of direct velocity measurements. To optimize this work and to specify its accuracy, we should be able to interpret the velocity discrepancy in terms of petrophysical parameter (e,g. density, porosity, etc. ), but these parameters for the two limestones have not yet been studied. S. Tillard, J.-C. Dubois / Journal of Applied Geophysics 33 (1995) 77-91 "x x x ~x Xx xXXXX x z x Xx x x xx KX X x x XXxx x x xx 1 XX .M. ' Xx 3 X Xx ~ x i 20 100 2OO 30O4O0 Frequency (MHz) Fig. 11. D i e l e c t r i c p e r m i t t i v i t y m e a s u r e d on s c h i s t s a m p l e s . + + + = sample shaped perpendicular to the schistosity, - - = sample shaped parallel to the schistosity. Although this interpretation is applicable to terrains favourable to electromagnetic propagation, its practical use is questionable in electrically conductive formations. Indeed, in this case, even if propagation is possible, it is generally very difficult to carry out CMP measurements since the antenna radiation pattern is more directive (Annan et al., 1975), making direct horizontal propagation measurements difficult. Moreover, slope calculation on steep time--distance curves tend to have a greater uncertainty. Applying this methodology to terrains with high velocities can also raise problems. In this case, the uncertainty arises from the difficulty encountered in separating the direct airborne wave from the direct underground wave. This mainly concerns glaciers in which the superficial layers are aerated. In an experiment on such a formation, we calculated values that reached 220 + 20 m / p s . Case history data corresponding to the different propagation velocities mentioned are shown in Fig. 10. We have referred to "superficial layers" without indicating the thickness parameters. In this regard, it is important to know what portion of terrain may be affected by direct propagation, since the wavelengths 89 involved in most radar surveys are of the order of one meter. This question is indirectly addressed in the study presented below where we investigated whether or not data recorded on an asphalt road over a sand dune modified the soundings significantly enough to prevent the accurate determination of the depth of the piezometric level under the dune ( -- 15 m). The road surface was made of asphalt, ballast (25 cm), and of a waterdraining synthetic textile coating. CMP measurements performed at two different locations of the dune on the road surface provided direct underground wave velocities of 141 ___4 and 139 + 2 m//xs. We then calculated the depths in time to the fresh water level using data collected from three available boreholes along the profile resulting in an average vertical velocity of 140 m / /zs. Because of the satisfactory agreement of the experimental and calculated data, we concluded that there is a non-quantifiable influence of the superficial asphalt road layer described above on the velocity parameter. An identical conclusion was put forward after comparing velocities measured in a granitic formation in a mine gallery, at a location where the ground was covered by a concrete surface 20 cm thick and where transmission measurements were recorded along the walls of a pillar. 3.2. Propagation velocity variation in case of anisotropy Among the many questions associated with radar surveys intended to provide geological information, we note in particular the problem of polarization rotation of electromagnetic waves in an anisotropic medium. In a survey relevant to this question (Tillard, 1991), data were recorded for four relative positions of the transmitter/receiver antennas along the profile (antennas first both perpendicular then both colinear to the line of measurement, and finally with the transmitter perpendicular to the receiver, the latter being first colinear and then perpendicular to the line of measurement). To simulate intermediate positions (any angle of the antennas with respect with the line of measurement) using these four basic configurations, we implemented software designed by the French Petroleum Institute for the study of shear wave birefrigency (Alford, 1986). In such anisotropy tests, we have to appraise attenuation, reflectivity, and velocity variations among the 90 S. Tillard, J.-C. Dubois / Journal ~'Applied Geophysics 33 ( 1995~ 77- 91 Direction of schistosity _____~R~ 105 m T = Trarmmller R = Receiver 14 m R R T Cross-cut Gallery Distance (m) 1 to O, 0 160' 160 320 ~ 320, 1 I lO i~ [.-, Fig. 12. 100 MHz transmission data recorded in a cross-cut ( l e f t ) and in a gallery (right) in an underground schist quarry with the antenna positioned vertically against the walls. various antenna configurations. In this paper, we report only the information on the velocity parameter measured in an underground schist quarry. Laboratory measurements illustrating the frequency dependence of both the real and imaginary components of the dielectric permittivity of rock samples are shown in Fig. 1 1. We complete the test with transmission radar data recorded in a gallery and in a cross-cut, parallel and perpendicular to the schistosity direction, respectively. We observe that the anisotropy in the schist formation leads to significant dispersion in the electromagnetic signals and variations in velocity values that depend on the schistosity direction considered. The radar records obtained in the gallery indicate a velocity of 100 + 5 m//xs whereas the velocity in the cross-cut was found to be 5 0 + 10 m//.ts at a frequency of 100 MHz. The uncertainties are higher in the case of a wave path perpendicular to the schistosity because of the higher attenuation when the electric field is orientated parallel to the schist lamina as shown in Fig. 12. Despite the S. Tillard, J.- C. Dubois / Journal of Applied Geophysics 33 (1995) 77-91 large margins of error specified in the velocity values, the impact of schistosity on the magnitudes of the velocity values is obvious. In this paper, the presentation of a few results concerning the effects of anisotropy empasizes the fact that the attention must be given to the interpretation of radar experiments carried out at sites where a geometric anisotropy, even slight, is likely to lead to a non-uniform velocity distribution. 4. Conclusion By means of the GPR field experiments reported here, several methods for determining propagation velocity in geological media have been presented. We are interested in the velocity because we must quantify the GPR performance in terms of investigation depth and reflector localization in the corresponding depth interval. In addition, the analysis of velocity in terms of petrophysical parameters will lead to a better knowledge of the propagation environment. The knowledge thus gained is more representative of the whole formation than that deduced from laboratory measurements of rock samples. In most cases, determining velocity is not an easy task and although it is possible to apply seismic migration processing to radar data, establishing a velocity profile law vs. depth is difficult. We have shown that only repetitive measurements in the field allowed calculation uncertainties to be reduced, but this procedure is insufficient to ensure errors of less than 10% when using velocity analysis with NMO corrections. 91 Finally, it is often difficult to determine the influence of certain information parameters on the calculated velocity. However, recording transmission measurements in various geographical directions may help to improve the wave propagation velocity assessment, and thus increase the quality of GPR surveys intended to provide accurate interpretation results. References Annan, A.P., Waller, W.M., Strangway, D.W., Rossiter, J.R., Redman, J.D. and Watts, R.D., 1975. The electromagnetic response of a low-loss, 2-layer, dielectric earth for horizontal electric dipole excitation. Geophysics, 40( 2): 285-298. Alford, R.M., 1986. Shear data in the presence of azimuthal anisotropy. 56th Soc. Explor. Geophys. Meet., Houston, TX, Expanded Abstr. Clough, J.W., 1976. Electromagnetic lateral waves observed by earth-sounding radar. Geophysics, 41 (6A): 1 126-1132. Coutanceau, N., 1989. Propri&6s di61ectriques de roches non argileuses dans la bande 20-1000 MHz, m6thode de mesure sur 6prouvette de dimension centim6trique. Thesis. Univ. Paris VII, Paris, 194 pp. Dubois, J.-C., 1992. GPR analysis in limestones. 54th Eur. Assoc. Explor. Geophys. Meet., Paris, Abstr. Fisher, E., McMehan, G.A. and Annan, A.P., 1992. Acquisition and processing of wide-aperture ground-penetrating radar profiles. Geophysics, 57 (3): 495-504. Fisher, E., McMehan, G.A., Annan, A.P. and Cosway, S.W., 1992. Examples of reverse-time migration of single-channel, groundpenetrating radar profiles. Geophysics, 57(4): 577-586. Shen, L.C., 1985. Problems in dielectric constant logging and possible routes to their solution. Log Anal., Nov/Dec: 14-25. Tillard, S., 1991. Evaluation of the performance of radar technique in geophysical prospection. Thesis. Univ. Strasbourg I, Strasbourg, 212 pp. Yilmaz, 0., 1988, Seismic data processing. Soc. Explor. Geophys., 2, 526 pp.
© Copyright 2026 Paperzz