Three-dimensional imaging of a deep marine channellevee/overbank sandstone behind outcrop with EMI and GPR

Transcription

1 THE METER READER Three-dimensional imaging of a deep marine channellevee/overbank sandstone behind outcrop with EMI and GPR RYAN P. STEPLER, ALAN J. WITTEN, and ROGER M. SLATT, University of Oklahoma, Norman, U.S. Deepwater channel-levee/overbank sandstones are important sources of oil and gas, but their internal stratigraphy is complex and not well understood. One means to better characterize this complex stratigraphy is through the study of outcrops, such as the Dad Sandstone member of the Lewis shale (Figure 1) in Carbon County, Wyoming, U.S. (Pyles and Slatt, 2000). Channel-levee/overbank systems are a result of multiple sediment gravity flows. The channel itself is formed from aggradation of levees over long periods of time. As sediment flows within the channel, fine particles are expelled over the channel edges, thus forming the levees. Coarser particles not deposited on the levee are transported beyond the channel to form frontal splays or lobes. Sediment may slump off the levee walls onto the channel floor during this time interval. Channels are later backfilled with sand and/or mud as transport energy diminishes, probably during an early rise in sea level. The materials that fill the channel can be configured in different arrangements of sedimentary structure and grain size. The channels are typically sinuous in nature, and appear similar to meandering river channels on seismic horizon slices and on the modern ocean floor. Previous ground-penetrating radar (GPR) work behind the Dad Sandstone channel-levee/overbank outcrop hints at the sinuosity of the channels, as does the distribution within the channel-fill sandstone of apparent cut bank slump and debris flow beds, and point bar cross-stratified sandstones. Also, certain sandstones appear to trend into the outcrop in an apparently sinuous manner. This study focuses on evaluating electromagnetic induction (EMI) as a tool for imaging such channels in the near surface. Outcrop mapping provides useful information regarding exposed boundaries and sedimentary grain size and structure; however, it only provides curvilinear two-dimensional geologic information. In order to obtain a complete 3D image of a channel-levee/overbank sandstone system, it is necessary to image the system behind the outcrop, where it dips beneath the ground surface. This can be accomplished in several ways: cores, well logs, and high-resolution, near-surface geophysics. Drilling and geophysics are difficult at the Dad Sandstone site due to steep terrain and coarse vegetation. GPR studies and limited drilling have been ongoing at the Dad Sandstone, but knowledge of the 3D structure has been evolving slowly due to difficulties in data acquisition. Electromagnetic induction was tested at this site as a rapid means to characterize lateral and basal geologic boundaries and trends. As discussed below, EMI responds primarily to the electrical conductivity of the subsurface and, 974 THE LEADING EDGE OCTOBER 2004 Figure 1. (top) The outcrop of the Dad Sandstone member of the Lewis Shale. A series of channelfill sandstones dip toward the lower right (west). The lowermost sandstone is the subject of this study. This sandstone trends along the outcrop face toward the lower right, but is juxtaposed against shale toward the west, at the same stratigraphic level. The GPR line on the lower left of the figure (modified from Young et al., 2003) shows the base of the channel-fill sandstone (red line) where it intersects the ground surface. Northwest of the base of the sandstone, at equivalent radar times, the lithology is shale. The lower right figure shows the EMI volume in plan view, with sandstone in yellow and shale in gray; note the bend in the sandstone at the location where the channel base intersects the ground surface. as such, should be capable of differentiating lower conductivity sandstone from higher conductivity shale. What makes EMI attractive is that it is light, portable, and does not require sensor contact with the ground surface. At the study site, EMI data can be acquired wherever it is possible to walk. In an initial evaluation of EMI in the Dad Sandstone, data sets were acquired adjacent to several outcrops. These results were quite encouraging, providing a qualitative picture that is consistent with exposed geology. In particular, as shown in Figure 1, the EMI data clearly display an area of low conductivity that exhibits the character of a sinuous channel sandstone. While these results might suggest that EMI can be used to map lateral sandstone/shale boundaries, additional and more rigorous studies are needed because EMI measures a volume-averaged response where this volume depends on the instrument s operating frequency. Thus, the information in Figure 1 may be interpreted as the structure at some shallow depth when it is, in fact, a response over some depth interval. This suggests that the sinuosity evident in the data may be an artifact associated with a more complex sandstone distribution. To better understand the implications of EMI measurements in defining channel sandstones, a second field study was performed east of the area shown in Figure 1. This area is adjacent to a well-displayed sandstone, is relatively flat and free of vegetation, and accessible for drilling. At this site, broadband EMI and GPR data were acquired. A portion of the EMI survey area was suitable for the acquisition of 3D GPR data that could verify and supplement the EMI information. In addition, a number of wells were drilled, and logged within and around the survey area. Geophysical methods. The subsurface can conduct electrical currents but the materials in the minerals that make up

2 Figure 2. Map of EMI and GPR survey areas. The EMI survey is the entire m area. The GPR survey area is confined to the gray box. Boreholes that have gamma logs are indicated, as is the relative location of the outcrop of channel-fill sandstone, and the strike and dip of the outcrop. This map also shows the location of the cross-sections in Figure 5. The cross-section numbers correspond to the borehole they intersect. the rocks in the subsurface, and more importantly the fluids within the pores, affect conductivity. In the absence of highly conductive objects, EMI can be used as a gross lithologic discriminator. However, EMI is seldom used in this manner, and rarely, if ever, in the study of sandstones in and behind outcrops. In this study, a volume of EMI data collected over a grid was used to map the 3D extent of the channel-fill sandstone (Figure 2). Sandstone dominates the channel-fill, while the surrounding levee/overbank facies are thin-bedded and shalier. Therefore, differentiating sandstones from shales in the EMI data is paramount to documenting the near-subsurface geometry of the leveed channel. The volume of multifrequency (1-20 khz) EMI data was collected over a grid adjacent to the outcrop of the channelfill sandstone. This volume was then inverted for resistivity (the inverse of electrical conductivity) and depth with software, based on an inversion algorithm by Huang and Fraser (1978). This algorithm is one-dimensional and assumes a horizontally layered half-space. For each measurement location, a model was defined as a series of horizontal layers and each layer has a magnetic permeability, electrical resistivity, and thickness. For this study, a 20-layer model was employed since 20 frequencies were used to collect the data. Twenty layers will allow for the maximum amount of variability in resistivity with depth. If fewer than 20 distinct resistive layers are present, the appropriate layers will have a thickness of zero and the resistivity values would be distributed over fewer layers. Other algorithms were used to give each layer an initial resistivity and thickness, and magnetic permeability. These initial values were then refined over a number of iterations. The layer parameters from the final iteration were used to create a synthetic EMI response that is then compared to the measured EMI response to create a misfit value defined as the percent difference between the synthesized response and the measured response. The inversion also estimates magnetic permeability for use in the resistivity and depth calculations since magnetic permeability has been shown to affect resistivity and depth calculations in previous studies. The inversion allows construction of a smoothed resistivity volume; however, it is important to note that vertical resolution will depend on the frequencies employed in the data acquisition. The inversion process is a mapping between skin depth and actual depth where, for a constant magnetic permeability, skin depth is proportional to the square root of the ratio of resistivity to frequency, so that higher frequencies Figure 3. A comparison of the gamma log, resistivity values from the inversion, and resistivity values after the interpolation at borehole 3. Differences between the two resistivity curves are attributed to the extraction performed to be as close to the borehole as possible, not each other. Included is a schematic stratigraphic column of the lithologic interpretation at the borehole to demonstrate the variability of lithology within the upper layer. are associated with smaller skin depths. Along with the EMI data, a volume of GPR data was collected over a small portion of the EMI survey grid. EMI data have been used to quickly ascertain the best place to conduct a GPR survey, but in this study GPR data were used to help interpret the resistivity volume inverted from the EMI data. The reflections evident in the GPR data are due to variations in dielectric constant and electrical resistivity. The channel-fill sandstone within the survey area is underlain by less resistive shale that limits the penetration depth of GPR. For this reason, and because of the shallow resolution limits imposed by the 20-kHz maximum EMI operating frequency, the depth of penetration of the GPR data is too shallow to match with vertical variability in the resistivity volume, although lateral changes in the GPR and resistivity volumes do coincide. This enabled a more accurate interpretation of the contact between the channel-fill sandstone and underlying shale. Five shallow subsurface gamma ray logs were also available within the survey grid; however, these were not deep enough to be directly tied to the resistivity volume. A direct tie would allow for a more exact lithologic interpretation of the resistivity volume. These logs did allow a second means of interpreting the resistivity volume which was accomplished through a comparison of sandstone content from the gamma log interpretation (based on well bore cuttings) with OCTOBER 2004 THE LEADING EDGE 975

3 Table 1. Percentage of lithologies at each borehole* Borehole # % Sandstone % Shale-Clast Conglomerate Figure 4. The result of plotting sandstone percentages against upper layer apparent resistivity values. The blue line is a power regression extrapolated to all resistivity values within the resistivity volume. The equation for the curve, and R 2 value are shown. upper portions of the resistivity volume (Figure 3). % Shale *Contains the percentage of the lithologies present at each borehole based on interpretations from Van Dyke (2003). Only the portion of the gamma log within the upper layer is used for the calculation of sandstone percentage at borehole 5. Data collection. A Mala GeoScience RAMAC/GPR system with 100-MHz center frequency antennae was used to collect the GPR data. Ten lines were collected, each separated by 1 m. Each line started at the western boundary of the area and extended east for 70 m. Each consecutive line was 1 m north of the previous line and walked in the same direction. The antennae were rigidly fixed 1-m apart and oriented perpendicular to the collection line. This configuration is considered a collocated multimonostatic collection geometry. Collection of the GPR data along 10 lines produced 10 two-dimensional data sets. The regular spacing of the lines also allowed the data to form a 3D volume. Processing each of the 10 lines was performed with a 2D imaging algorithm based on geophysical diffraction tomography. The instrument used to collect all EMI data was the Geophex GEM-2 which has a coil separation of 1.67 m and the capacity to function with 10 frequencies at a time. In order to collect the 20 frequencies desired for this study, an initial group of 10 frequencies was collected over the survey area and then the survey was repeated with the other 10 frequencies. The resulting 20 frequencies were grouped from 1-10 khz and khz. For simplicity of reference, the frequencies employed are rounded to khz; however, the actual acquired frequencies are at uniform intervals between 990 Hz and Hz. The collection grid is just east of where the channel-fill sandstone crops out. A m grid was marked in this area such that the long axis runs east-west, and the short axis north-south. Data were collected along paths in the east-west direction, changing direction after each line. Lines were separated by 2 m in the north-south direction. The previously described GPR grid lies within this larger EMI grid and shares the eastern, western, and southern boundaries. The GEM-2 acquires data at a rate of about five samples per second while the instrument is being continuously walked from one grid edge to the other. Thus, the number of data points per line varied depending on the speed at which the operator was walking. The software for download of the instrument interpolates the data onto regular intervals along each line walked for the two data groups. The two datasets do not have the same interval spacing due to inconsistent walking speeds between the two groups. This problem was successfully resolved during processing. Interpretation. The final interpretation is the integrated result of analyzing three data sets: EMI, GPR, and samplecalibrated gamma ray logs. EMI is the focus of this study and ideally it would be sufficiently detailed to interpret it alone. This is not the case, and therefore the other methods were utilized. The values obtained from the layered model inversion are apparent resistivity values of that layer, not true resistivity values. The resistivity values returned by the inversion algorithm are averages over the thickness of each layer and can thus constitute a composite of various materials within a given depth interval. It is important to recognize that a layer identified as sandstone may have a higher (apparent) resistivity than pure sandstone as a result of a slight increase in moisture or shale content. Any reference to resistivity in regard to the inversion results is actually a reference to apparent resistivity. The apparent resistivity values are used to enhance the lithologic interpretation of the resistivity volume and extracted cross-sections. A few assumptions are made about mineral assemblages and fluids within the survey area, as resistivity is not typically used as a lithologic discriminator. Within this small study area, the distribution of minerals within the sandstones and shales is assumed homogeneous. The second assumption is that the fluid composition is assumed constant. The water table, at a depth of approximately 60 m, is too deep to be detected by this study. A sharp boundary is present in cross-sections of the resistivity volume inverted from the EMI data, but is not thought to be attributed to the presence of a perched water table, and will be discussed later. The mineral and fluid assumptions leave only lithology as the variable that affects the EMI data. Figure 2 shows the positions of the five boreholes within the EMI and GPR survey grids. Four boreholes are no deeper than the upper layer inverted from the EMI data at the same location. Figure 3 shows that at borehole 3 a direct tie between the gamma ray logs and the extracted resistivity values cannot be made because of a thick top model layer. This is true for all boreholes. If lithologic changes occur within this layer, they cannot be resolved by our EMI measurements. Consequently, the imaged resistivity is a depthaveraged value over all materials present and the vertical structure cannot be resolved. For this reason and because most boreholes are no deeper than the EMI upper layer thickness, it is impossible to directly correlate either the gamma logs or their interpreted lithology with inverted resistivity as a function of depth. In order to use resistivity from the EMI inversion as a diagnostic for lithology, the depth-averaged value of resistivity in the uppermost layer in the vicinity of each borehole was correlated with a depth-averaged percent sandstone over a comparable borehole depth (Table 1). Figure 4 presents a plot of percent sandstone versus resis- 976 THE LEADING EDGE OCTOBER 2004

4 Figure 5. (a) The two resistivity cross-sections with axes in meters that intersect borehole 2. The color bar gives the resistivity values, and percentages of sandstone present. Sandstone percentage correlated to resistivity is determined from Figure 4. The small brightly colored cross-section is extracted from the GPR volume where there is overlap with the presented resistivity cross-sections and is at the same scale as the resistivity cross-sections. Colors in the GPR cross-section are for contrast enhancement only. (b) The same as (a) but for cross-sections that intersect at borehole 3. There is also another GPR crosssection because of more overlap. The final new features are the channel base interpretations, in black on the GPR cross-sections, and superimposed in red on the resistivity cross-sections. The red and black lines are extended above, and out of, the cross-sections for visualization purposes only. (c) This figure is similar to Figure 5b except these cross-sections intersect at borehole 6. tivity for the five boreholes and a power law curve-fit based on a regression analysis. It is important to recognize that the power law is extended to resistivity values and percent sandstone that are below any measured values and that, where data are available, the fit is relatively good except for borehole 5 where the percent sandstone is 78% (Table 1). This outlier could be a result of greater shale content within the depth interval considered, as suggested by the irregular character of the gamma log. While unreliable for low resistivity values, the curve fitting processes does suggest a viable methodology for lithologic interpretation for EMI measurements that can more reliably be employed in the future when more gamma log information is available. The base of the channel-fill sandstone is contained within the gamma log at all locations except for boreholes 5 and 2, where the base of the channel-fill sandstone is at the bottom of the borehole. Once again, the upper layer in the resistivity volume is too thick to show any variability at the channel base. For this reason, inverted resistivity from the EMI data along with lithologic ties established from the regression analysis can be used to define lateral, but not vertical, boundaries of the channel sandstone. Therefore, the GPR data were interpreted for the channel base starting at borehole locations. These interpretations were then overlain on the resistivity cross-sections to determine a resistivity value that can be associated with the lateral extent of the channel. Figure 2 shows the location of six cross-sections (Figures 5a-c) taken through the resistivity and GPR volumes to intersect borehole locations. Where an interpretation is made on the GPR data, it is overlain on the resistivity data. Based on the GPR interpretation, we chose a value of 200 ohm m as the cutoff value for the lateral extent of the channel. From Figure 4, this apparent resistivity value is associated with an approximately 62% sandstone content. Within the channel-fill sandstone, the near-surface resistivity is high and this layer is relatively thick. Outside of the sandstone, greater shale content reduces the resistivity of this layer. As a result, the inverted thickness of the uppermost layer should diminish outside of the channel. This expectation is consistent with the GPR interpretation of channel boundaries in Figures 5b-c. The sharp horizontal boundary in the resistivity cross-sections is interpreted not to be the base of the channel for the reasons given above. It is also not interpreted to be a locally perched water table. From the gamma logs, and previous studies, it is known that a thick shale layer exists below the channel sandstone. Shale does not typically contain aquifers, and this increases the confidence in a lack of a perched water table. The other arguments against the presence of a perched water table within the resistivity volume are that the boundary does not follow the topography consistently and that groundwater was not encountered in the wells. This is especially evident in cross-sections that run north-south. The sharp boundary is believed to be the result of the upper layer containing predominately sandstone, while lower layers contain a high percentage of shale. The sizeable thickness of the upper layer throughout the volume does not allow for a gradational change, or correct boundary location, and therefore creates the sharp boundary present in the volume. In a depth slice through the resistivity volume at 2 OCTOBER 2004 THE LEADING EDGE 977

5 Figure 6. This is a depth slice through the resistivity volume at two meters depth with the axes in meters. The outline of the channel is in red, with the purple being interpreted as a transition zone discussed in the text. m, the lateral extent of the channel defined by 200 ohm-m is outlined in red (Figure 6). This red line bounds the channel in the southwest corner of the depth slice. Although this outline is jagged, curvature can be seen in the channel supporting the interpretation of a sinuous bend in the channelfill sandstone. In the southeastern corner of the depth slice a small area is outlined, containing the same resistivity values as the channel-fill sandstone. This feature is interpreted to be a splay deposit from the channel margin. A feature like this would not be unusual on the cut bank side of a channel at a bend. It is also believed that the flow within the channel was from the west, supporting the idea of a splay. There is also a band (purple) which is a transition zone that may contain the actual channel margin. Although the lateral extent of the channel is well defined by the resistivity data, it is difficult to determine what is occurring geologically within the transition zone. The decrease in resistivity across this zone could be a result of slumped beds on the cut bank side of the channel. Slump material at a cut bank is noted at outcrops of the same channel sandstone outside this survey area. Conclusions. EMI as a tool to study channel-fill and other sandstones in outcrop gives a good lateral picture of the channel margins and trends where the beds dip beneath the ground surface. Unfortunately, the base of the channel is not accurately imaged and is a manifestation of the upper layer thickness that was obtained in the inversion. The 3D image obtained from the resistivity volume lacks the gentle slope that occurs between the channel center and channel margins as depicted in the GPR data. The image also fails to depict the correct vertical position of the channel base as defined by the gamma ray logs and GPR profiles. Accurately imaging the base of the channel may be accomplished by using higher frequencies in acquisition of the EMI data. Also, boreholes drilled within an EMI survey site for comparison to the EMI inversion results should be deeper, such that a direct tie between the two types of data can be made. Nevertheless, EMI has an advantage over GPR for this situation. EMI data are collected much more quickly, and can cover ground that would have to be cleared of vegetation for GPR acquisition. Antennae used in GPR are relatively narrow but have lengths that are proportional to their center frequency. The 100-MHz center frequency antennae used in this study are about 1 m long. In order to avoid substantial incident energy loss that results from reflections at the ground surface, the antennae must rest on or be only slightly elevated (no more than a few centimeters) above the ground surface. In areas of dense (even low-growing) vegetation, it may be necessary to either remove the vegetation or, if possible, to insert the antennae under the vegetation canopy. Such procedures can greatly increase the level of time and effort required in GPR data acquisition. If geologic information can be tied to the EMI inversion results with little or no GPR data, lithologic and geometric information can be interpreted easily throughout an entire volume with a greatly reduced field effort. A new version of the GEM-2 EMI instrument offers a highest operating frequency that is more than twice that used in this study. This will allow better near-surface resolution. The EMI result in Figure 6 is suggestive of a channel bend. If this is the case, a bend could be better defined by extending the study to the south. For this reason, future geophysical studies are planned at this site with the new GEM-2. This effort will also include deeper wells for logging, and high-resolution seismic reflection to provide lithologic boundaries somewhat deeper than those that could be defined from the GPR measurements. Suggested reading. The differential parameter method for multifrequency airborne resistivity mapping by Huang and Fraser (GEOPHYSICS, 1996). Simultaneous reconstruction of 1D susceptibility and conductivity for electromagnetic data by Zhang and Oldenburg (GEOPHYSICS, 1999). A high-frequency sequence stratigraphic framework for shallow through deepwater deposits of the Lewis Shale and Fox Hills Sandstone, Great Divide and Washakie Basins, Wyoming by Pyles and Slatt (Gulf Coast Section, Society of Sedimentary Geology 20th Annual Research Conference, 2000). Application of groundpenetrating radar imaging to deepwater (turbidite) outcrops by Young et al. (Marine and Petroleum Geology, 2003). A fine scale 3D architecture of a deepwater channel complex, Carbon County, south-central Wyoming by S.K. Van Dyke (unpublished M.S. thesis, University of Oklahoma). Acknowledgments: This work was supported by Shell International Exploration and Production and Conoco-Phillips Inc. We thank I.J. Won at Geophex for technical support and Staffan Van Dyke for the gamma log interpretations. Ryan Stepler is currently with Marathon Oil Company. Corresponding author: awitten@ou.edu 978 THE LEADING EDGE OCTOBER 2004