Download PDF version

Abstract

We acquired 15 km of seismic reflection and refraction profiles along a highway in central Illinois to locate old subsurface mine workings that may indicate areas of potential instability. Illinois Route 29, a major highway along the western side of the Illinois River Valley, swings close to steep bedrock bluffs for several miles in northern Peoria and southern Marshall Counties. Coal seams that crop out along these bluffs have been intensively mined and two mines, both now abandoned, were known to work deposits beneath the level of the valley floor. The seismic investigation was part of a detailed geologic study of the highway corridor in preparation for a major renovation project.

The 48-channel data set was collected using a 100-lb accelerated weight drop source with a 20-ft source spacing and 10-ft receiver spacing. We applied standard reflection processing and a simple 2-layer velocity model for refraction analysis. Disruptions in the bedrock reflections at the depth of the mines were imaged at several locations. Refraction analysis indicated irregularities on the bedrock surface that may be caused by collapse or landslide. Integrating the two seismic techniques provides greater insight into the subsurface than either employed separately.

Introduction

Illinois Route 29 (IL 29) is a major north-south transportation route along the western side of the Illinois River Valley (Figure 1). For the most part, the wide river valley provides ample room for an adequate highway corridor. But for several miles north of the town of Chillicothe in northern Peoria and southern Marshall Counties, the river channel swings close to the western margins of the valley, forcing the highway against steep bedrock bluffs. The proximity of the bluffs presents two important geotechnical hazards, potential landslides and potential collapse from abandoned shallow coal mines. The Illinois Department of Transportation (IDOT) has requested assistance from the Illinois State Geological Survey (ISGS) in investigating these and other technical questions concerning a planned improvement of the IL 29 corridor in this area. The identification of underground mines is a critical component of the ISGS’s mapping activities along the IL 29 corridor. In this part of the project, geophysical techniques are used to locate old mine workings that may be present beneath the highway corridor. The technique can also estimate the size and depth of the underground mines and indicate areas of potential landslide activity.

Geologic background

Near-surface Quaternary Geology

Quaternary sediments in this area are found both at the top of the river bluffs and at the ground surface, at the base of the bluffs. The bluffs, which stand about 200 ft high, are capped by up to 5 ft of Peoria Silt, a late Wisconsin Episode, wind blown loess (Hansel and Johnson, 1996). Beneath the loess are about 40 ft of sandy diamicton and sand lenses of the Tiskilwa Formation, Wedron Group (Hansel and Johnson, 1996). These glacial sediments were deposited during the Wisconsin Glacial Episode. Beneath the roadway, at the margin of the Illinois River Valley, Quaternary and Recent sediments range from less than 10 to over 30 ft thick. Most of these materials are alluvial and colluvial sediments of the Cahokia Formation. Thicker sediments of the Henry and Equality Formations, which were deposited during the Wisconsin Glacial Episode (Hansel and Johnson, 1996), may be present at the north and south ends of the study area where the roadway leaves the margin of the valley.

Paleozoic Bedrock Geology

Most of the bluff material consists of sedimentary rocks of the Modesto and Carbondale Formations of Pennsylvanian age (Smith and Berggren, 1963). Undercutting of the Lonsdale Limestone Member of the Modesto Formation, which occurs near the top of the bluff, has resulted in large blocks of limestone in the sediments near the base of the bluffs. The Danville (No. 7) Coal Member, the uppermost member of the Carbondale Fm. occurs near the base of the bluff. In places this coal bed is 3.5 ft thick and was the target of many small surface and drift mines. Just above the Danville Coal is the Farmington Shale Member which was mined for brick production at a large facility in the southern part of the study area (White and Lamar, 1960).

The cyclic sedimentary rocks of the Carbondale, Spoon, and Abbot Formations are up to 300 ft thick beneath the study area. Most of these sediments are shale, but thin limestone and variable sandstone layers provide important stratigraphic markers and seismic reflectors. The coal members in this area are, for the most part, thin and poorly developed. Only one subsurface coal mine is known from the study area that is not associated with bluff-side drifts. This mine used the long-wall method to extract the relatively thin (30 inches) Colchester (No. 2) Coal at a depth of about 180 ft beneath the town of Sparland (Smith and Berggren, 1963). One of our seismic lines skirts the edge of this mined area.

When the brick company had exhausted the Farmington Shale pits, it began to mine clay with a room and pillar operation at a depth of about 250 ft. This mine extracted an 8 to 10 ft thick clay layer near the contact between the Spoon and Abbot Fms. Mine maps indicate that this operation extended beneath the study area for about 0.5 miles. Our Borehole #3 was located near now abandoned surface structures from this facility. Drilling in this borehole ceased when a void was encountered at a depth of 251 ft, presumably in the abandoned mine works.

An unconformity separates the Abbot Fm from the distinctive dark and variegated shales of the New Albany Fm (Devonian) (Willman et al., 1975). These shales are 200 to 300 ft thick in this area. The unconformity is significant because thick carbonate rocks of Mississippian age usually intervene between the Abbot and the New Albany. This accounts for the absence of the pronounced seismic reflection which would be expected from the carbonate rocks in our study area.

Methodology

Theoretically, underground voids, such as mine workings or collapsed mine workings, exhibit very strong impedance contrasts with surrounding rock and have good potential for being observed by the reflection method. The impedance contrast of Quaternary sediments or weathered bedrock (both having low velocity and density) with unweathered bedrock (having high velocity and density) is also very strong. In practice, the complex geometry of the voids often makes them difficult to distinguish from background noise in the data. Also, irregularities in the shallow sediments (thickness and material type) introduce strong distortions in the reflection images that may lead to erroneous interpretations of voids where none exist. Special attention has been given to the correction of this problem.

During the processing and interpretation phases we added a special refraction interpretation routine similar to refraction statics corrections. This process resulted not only in the static correction, but in a simple, 2-layer model of the shallow subsurface. We used this model to explain several irregularities in the reflection data, and it can also be used as an engineering tool for road design. In particular, where the refraction model indicated that the bedrock was very shallow (less than 10 ft beneath the ground surface) the reflection data frequently contained strong, spurious coherent noise, known as guided waves (Robertson et al., 1995). We were only partly successful in removing this noise from the data set, so having the refraction data as an independent check helped to confirm where signal degradation was caused by actual subsurface conditions or was an artifact of the surface conditions.

Preliminary tests across known underground coal and clay mines revealed that standard refraction methods are not compatible with the subsurface conditions at the site. Signal variations attributable to the possible presence of collapsing structures cannot be easily separated from signal variations caused by changes in bedrock depth and other subsurface variations. However, the same tests showed that seismic reflection methods would give good results.

Data acquisition

We acquired 9-½ miles (15.2 km) of seismic data in less than 8 working days, see Figure 1.With the assistance of IDOT District personnel for traffic control, data acquisition proceeded at a rate of more than 1 mile/day (1.6 km/day) with a shot spacing of 20 ft (6 m) and a receiver spacing of 10 ft (3 m) (see table 1 for the recording parameters). The energy for the system was derived from the impact of an accelerated 100-lb weight dropped onto the pavement. The energy pulse is transmitted to the ground beneath the road through an aluminum plate. Initial testing indicated that the weight bounced on concrete pavement, producing a second pulse about 100 ms after the initial impact. The second pulse caused considerable noise in the data, often obliterating the true signal. We found that we could dampen the bouncing of the weight by inserting a lead plate beneath the aluminum impact plate. This adaptation did not affect the initial energy pulse, but the deformation of the lead consistently dampened the bouncing of the weight and eliminated the second pulse.

Figure 2 shows three raw shot records chosen from the survey to illustrate some of the processing issues that had to be addressed in this study. The first record was acquired above the old 80 m deep clay mine. Deep reflections are almost missing and major static shifts are present in this record. This time shift is corrected by alignment of the head waves (Fig. 2.C). The other two records illustrate the extreme variation in signal frequency obtained in this survey. High frequencies present in Record 2 are missing in Record 3. Surprisingly, these two shots are separated by only 20 ft (6 m). For Record 2, the thumper was located on a concrete bridge and for Record 3 the shot was on the artificial fill next to the bridge. Apparently, at the location of Record 2, the sound propagated through the foundation of the bridge which was set on the bedrock itself. The lower frequencies on Record 3 are caused by a natural high-cut filter effect induced by the shallow unconsolidated artificial fill. A band-pass filter of 110 Hz-165 Hz applied to both records removes these differences (Fig. 2.B). Records 2 and 3 illustrate another common problem with this data set: a low-frequency, high-velocity guided wave. For these two records a band-pass filter adequately removes both the guided waves and the ground-roll. However, when the bedrock was shallower than 5 ft (2 m) we were not able to remove the high-frequency guided wave phase from the data set. This coherent noise has slightly decreased the quality of the section in those areas.

Data processing

Reflection processing

Seismic reflection processing was performed using PC-based software and, for the most part, followed standard processing protocols. Because of the severe shallow distortions, we applied our own routines for static (time) correction analysis and modeling (operations 3, 4 and 8 in table 2, for further information see Pugin and Pullan, (2000)). This near-surface correction process involves a very careful measurement of the initial arrival of the seismic signal (the first break) at each of the 48 channels recorded after each of the 10,250 impacts. The time shift, which is mainly a result of variations in the thickness of the unconsolidated sediment, severely affects the quality of the stacked seismic section. Figure 3 shows how removal of these time shifts through the alignment process tremendously increases the quality of the section. The section displayed in Figure 3 shows an anticlinal feature with fractures on its left edge. The large amplitude observed in the envelope attribute section and the repeated or “multiple” energy present below the strong reflection implies that this anticline or dome may contain methane.

Receiver-based refraction processing

For nearly all the records in this study, the initial energy recorded at the receivers is either the direct energy wave traveling through the ground to the receiver, or energy which has traveled down to the top of the bedrock and then refracted back to the ground surface (Figure 3). For this reason, the first breaks, which were carefully measured for reflection static correction, can also be used for refraction analysis of the near surface. We developed a simple refraction analysis method to take advantage of this large amount of near-surface data. This analysis provides an estimate of the thickness of the near-surface materials (soil and weathered bedrock) every 10 ft (3 m) along the survey line. It is designed for the specific conditions along IL 29 and has the following restrictions:

1) Near-surface materials (soil and weathered bedrock) are treated as a single shallow layer over bedrock;
2) The shallow layer is not thicker than approximately 30 ft

In most of the surveyed area, these assumptions are valid. One exception has been found on the most southern section, where two shallow layers are present above bedrock. Since mine voids are not observed or suspected in this short section, depth to bedrock has not been computed.

Description of the method

This method is based on the calculation of the variation of delay times generated by changes in the near- surface layer. We have simplified the analysis by assuming that all the variations are due to thickness variations in the near-surface layer. We have used a single, average velocity for sound propagation in the near-surface layer and a second, average velocity for sound propagation in the bedrock. Our refraction calculation requires several steps:

a) The associated first-break geometry is sorted in common-receiver records.

b) We choose first breaks that correspond to the second refracted layer. All offset data less than 120 ft are removed.

c) An average offset and first-break time is computed for each receiver. To be sure that trigger inaccuracies are filtered, only receivers with 16-fold or higher are used.

d) A time shift is applied to remove the offset effect: t0 = t – (a / 8500 ft/s or 2590 m/s) where t0 is the time at the zero distance offset, t is the average first break time, a is the average offset and 8500 ft/s or 2590 m/s (P-wave) is an estimated bedrock velocity propagation of sound generally observed on the data set.

e) Depth is calculated by using an average low-velocity (soil) of 1500 ft/s or 450 m/s (P-wave) and the equation: h = ½ * 1500 ft/s * t0 where h is the depth at the receiver location.

An example of first breaks from a series of shot records is shown in Figure 4, upper graph. Our refraction calculation, represented in the lower graph, estimates the depth of unconsolidated sediment overlying the bedrock. The shape of the lower curve matches that of the first breaks in the upper display. For example, a trough in the center of Figure 4 correlates with a delay in the head waves.

A comparison with existing borehole data shows that this method adequately estimates bedrock depth within an error of a few feet. However, the method does not account for lateral changes in the near-surface materials. Stream sediments, for instance, often exhibit relatively low propagation velocities compared to other shallow materials. Our analysis typically indicates thicker materials in stream valleys, but some of this apparent thickening may be a result of lower propagation velocities. Considering this observation, the depth to the top of bedrock may be over-estimated in stream valleys and other places where thicker soil layers may be present.

Borehole analysis, Seismic calibration and Horizon picking

Three boreholes were drilled to provide geologic control for this project. Data from one of these boreholes, borehole 3, is presented in Figure 5, location in Figure 1. This borehole was drilled in an area suspected of being undermined. It was located near the foundations of abandoned surface structures associated with the old clay mine. Drilling operations were halted when a void was encountered at a depth of 251 ft. According to mine notes on file at the ISGS, this is the depth of the mine.

Lithologic and stratigraphic analyses were based on core extracted from the borehole. A suite of geophysical logs provided correlation with the seismic reflection data. Several attempts were made to acquire a Vertical Seismic Profile (VSP) and a velocity log based on first breaks using a down hole geophone and a down hole multi-channel hydrophone array (provided by the Geological Survey of Canada, Ottawa). However, signal quality, particularly at this site, was so poor that first breaks could not be picked. We suspect that this is related to fracturing associated with the mine operation. Signal quality was somewhat better in borehole 2 (location in Figure 1) which was located in an unmined area. We were able to determine an average P-wave velocity of 8500 ft/s (2590 m/s) for the bedrock. This is the same value determined by the refraction analysis. The seismic reflection profiles were then converted to depth sections using the constant average velocity of 8500 ft/s and exported to the Kingdom Suite seismic reflection interpretation package.

Using the lithologic log and the electric logs, we were able to correlate strong reflections with the Hanover limestone (Figure 6, yellow horizon), the base of the Pleasantview sandstone, (Figure 6, light-blue horizon) and the base of the Isabel sandstone (Figure 6 green horizon). A fourth reflector is probably correlative to the Babylon sandstone or the Pennsylvanian-Devonian unconformity (Figure 6, red reflector).

Results

As an example of the results we achieved in this study, we present a 7000 ft (2.33 km) section of one of our seismic lines south of the village of Sparland (Figure 6). Mine maps indicate that the old clay mine underlies at least part of this section. In contrast with the section displayed in Figure 3, this section shows strong changes in amplitude and oblique reflections which can possibly be interpreted as diffractions. Some of these apparent diffractions collapse during the migration process. However, because many of the 3-dimensional subsurface structures causing these diffractions are off-line relative to the seismic line, the 2-dimensional migration algorithm is not successful in removing all the oblique noise. Slope stability has been an ongoing problem in this stretch of IL-29. A tied-back retaining wall was installed along a half-mile section in order to stabilize the base of the bluff. In this illustration, static corrections and filtering have been applied. Approximate depth scales are shown. Geophysical logs from Borehole 3 are shown for correlation. Borehole 3 was terminated at a depth of 251 ft when drilling operation encountered a void, presumably in the old clay mine. Incoherent noise that disturbs the reflections is present in several areas at that depth. We have circled areas where the noise is not associated with either surface features such as culverts or processing artifacts from guided waves. These areas are within the boundaries of the mapped clay mine and we believe they indicate voids in the bedrock. Several smaller diffractive features are also present in this section, but are not individually noted. Both the northern and southern edges of the mine appear to be independently imaged in this section.

Conclusions

After careful processing, using special software tools to remove near-surface delay-time effects, the seismic reflection method was used to map subsurface mines. Two criteria were employed to map the void structures: a decrease in reflection amplitude and the presence of diffractive reflection energy. The interpretation was improved by the use of electric logs, even though a down hole velocity calculation was not possible and no VSP was available. The data quality would have improved if the shot spacing were decreased from 20 ft (6 m) to 10 ft (3 m) so that the CMP fold were increased from 12 to 24. However, despite the choice of this wide shot interval, we were able to confirm the presence or absence of undermined areas.

Acknowledgements

This project was funded by the Illinois Department of Transportation, John Washburn, project manager. Mike Lewis, IDOT Peoria District, supervised field traffic control. Steve Sargent, Tim Young, Pius Wiebel and Andy Stumpf of the ISGS contributed to this effort. Seismic interpretation software was provided through an educational grant from Seismic Micro Technology, Inc.

References

Hansel, A.K., and Johnson, W. H., 1996. Wedron and Mason Groups : Lithostratigraphic Reclassification of Deposits of the Wisconsin Episode, Lake Michigan Lobe Area. Illinois State Geological Survey Bulletin 104. 116 p.

Pugin, A. and Pullan, S.E., 2000. First-Arrival alignment static corrections applied to shallow seismic reflection data. Journal of Environmental & Engineering Geophysics, 5/1, 7-15

Robertson, J. O. A., Pugin, A., Holliger, K. and Geen, A., 1995. Effects of near-surface waveguides on shallow seismic data. SEG Abstract, Houston, Texas.

Smith, W.H., and Berggren, D.J., 1963. Strippable coal reserves of Illinois, Part 5A: Fulton, Henry, Knox, Peoria, Stark, Tazewell, and Parts of Bureau, Marshall, Mercer, and Warren Counties. Illinois State Geological Survey Circular 348. 59p.

White, W.A., and Lamar, J.E., 1960. Ceramic tests of Illinois clays and shales. Illinois State Geological Survey Circular 303. 729p.

Willman, H.B., Atherton, E. Buschbach, T.C., Collinson, C., Frye, J.C., Hopkins, M.E., Lineback, J.A., and Simon., J.A., 1975. Handbook of Illinois stratigraphy. Illinois State Geological Survey Bulletin 95. 261 p.