Developing a Rock Mass Tilt and Seismic Observatory at DUSEL

Transcription

1 ARMA Developing a Rock Mass Tilt and Seismic Observatory at DUSEL Sherman, C. S., Magliocco, M., and Glaser, S.D. Dept. of Civil and Environmental Engineering, University of California Berkeley, Berkeley, CA, USA Copyright 2011 ARMA, American Rock Mechanics Association This paper was prepared for presentation at the 45 th US Rock Mechanics / Geomechanics Symposium held in San Francisco, CA, June 26 29, This paper was selected for presentation at the symposium by an ARMA Technical Program Committee based on a technical and critical review of the paper by a minimum of two technical reviewers. The material, as presented, does not necessarily reflect any position of ARMA, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of ARMA is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgement of where and by whom the paper was presented. ABSTRACT: The Transparent Earth group is developing a permanent seismic observatory at the Deep Underground Science and Engineering Laboratory (DUSEL). We are currently operating two stations on the 2000 level, and one on the 4100 level. Each seismic station has a triaxial borehole accelerometer and a biaxial floor-mount tiltmeter, which is capable of measuring long-period tilt as small as a 0.1 µradians. In situ calibration of the accelerometer array is determined through modeling of the 9.0 Mw 2011 Tohoku, Japan earthquake; filtering is used to remove the effect of local temperature variations on the tiltmeter data. The individual earth tide signals are estimated using a linear, elastic, solid earth tide model corrected for ocean tide loading, topographic effects, and the effect of lateral inhomogeneity in the nearby rock. The extracted O 1 and M 2 earth tide modes show significant diurnal and semi-diurnal components, respectively. The best-fit analytic model for these modes was used to estimate the bulk elastic properties local underlying crust, in particular rock mass compressibility. We assumed that the long-term trends in the tilt signal (approximately 3µradians/day towards the pumping well) are due to the compression of the underlying aquifer from the extraction of groundwater. Over the range of bulk hydraulic conductivities expected for the rockmass (between 10-6 and 10-7 cm/s), our simplified model of the aquifer suggests that the bulk storativity is between 10-3 and INTRODUCTION The Homestake gold mine, which is located in the Black Hills of South Dakota, is the deepest and longest operating gold mine in North America. During the operation of the mine from 1876 to 2001, approximately 1200 tonnes (42,000,000 oz) of gold was removed. The mine includes over 560 km of drifts and extends to a depth of over 2.4 km (8,000 ft). Due to low gold prices, mining operations ceased in In 2003 the mine was sealed and the groundwater pumps were turned off, allowing water to flood the lower levels of the complex. In July of 2007 the mine was selected as the location for the new Deep Underground Science and Engineering Laboratory (DUSEL), a project that will provide the infrastructure for a wide array of scientific experiments to advance the fields of physics, biology, earth science, and engineering. As part of the laboratory s early baseline monitoring efforts, the Transparent Earth team was allowed early access to the mine to install instruments to measure deformation and transient ground motions in the local rock mass due to the construction and dewatering of the laboratory. This effort was the first scientific project to take place in the facility since the reopening of the mine. Taking advantage of the thousands of open boreholes throughout the laboratory, we deployed the initial stage of a facility-wide ground deformation and seismic observatory at DUSEL [1]. Each Transparent Earth station consists of a high frequency, triaxial borehole accelerometer and a low-frequency, biaxial, platformmounted tiltmeter.

2 2. RESEARCH GOALS The primary goals of our project were as follows: 1. Estimate the realistic base noise level for tilt measurements and the degree to which measurements include transient mechanical and thermo-mechanical effects. 2. Develop deterministic and adaptive filters to remove transient signals. 3. Use low frequency, teleseismic signals to calibrate and orient the sensors in space. 4. Measure different earth tide constituents to estimate the bulk elastic properties of the crust. 5. Develop a simplified hydrologic model of the rock mass capable of estimating the tilt signal due to compression of the underlying aquifer. For analysis of the tilt data, we assumed that the measured signals are the superposition of three signals: First, thermo-mechanical tilt due to temperature effects from the instrument mounting and rock mass. Second, permanent tilt due to changes in the stress field from groundwater extraction, facility construction, and tectonic loading. Third, cyclical or transient mechanical tilt, due to atmospheric loading, earthquakes, and earth tides. 3. EQUIPMENT OVERVIEW We currently operate two stations that are measuring rock-mass tilt. The locations of the instruments are shown in Figure 1. The tiltmeters are Applied Geomechanics 711-2A platform-style tilt-meters, which are capable of detecting rotations as small as 0.1 µradians [2]. The data are sampled at a rate of 1 Hz, resulting in oversampling by a factor of 25, using a Diamond-MM-16-AT 16 bit A/D converter; the data are transmitted to the surface using the laboratory network. The signals are low-pass filtered to 0.1 Hz to isolate long wavelength components of seismic events, earth tides, seasonal trends, and stress realignment associated with mine dewatering and construction. Because our sampling rate (f sample =1 Hz) is much greater than the minimum Nyquist frequency (f nyq =0.04 Hz), our tilt measurements are not temporally aliased. The tiltmeter uses two orthogonal, electrolytic level sensors contained within a hermetically sealed enclosure mounted to a 15 cm by 15 cm anodized aluminum base. The tiltmeter assembly is bolted to a 30 cm diameter concrete pedestal reinforced by three 1.9 cm diameter rebar rods embedded at least 30 cm into the rock of the tunnel invert. Fig. 1. Plan view of DUSEL showing the location of currently installed tilt-sensors. 4. METHODOLOGY In this analysis, we attempted to deconvolve the effect of the different processes from the tilt signal. Our primary assumption was that over a range of time scales a single process dominates the tilt signal, and that these processes may be modeled as a linear time-invariant (LTI) system. Over very short time scales (minutes to several hours), the tilt signal was dominated by stationary temperature effects. For the longest time scales (months to years), the data was dominated by transient tilt due to the extraction of groundwater from the rockmass. Over intermediate time scales (days to months), the tilt signal was dominated by atmospheric loading. By modeling and removing these effects from the tilt data, we gain insight into the thermal, mechanical, and hydraulic properties of the rock mass. We also considered a variety of other signals in our analysis, such as teleseismic earthquakes and earth tides, in order to calibrate and orient our instruments and gain insight into the mechanical properties of the underlying earth s crust. For segments of this analysis, we found that it was convenient to consider the time-derivative of tilt using polar notation. The magnitude of the tilt rate (! t ) is given by Equation 1, where! 1 and! 2 are the tilt rates for sensors one and two. For very small tilt rates, the small angle approximation may be used to simplify the tangent terms.. The direction of maximum tilt in the horizontal plane is given in Equation 2, where α is the counterclockwise angle from sensor 1.

3 ( (! 1 ) + tan 2 (! ))!! 2! t = atan tan 2! ( ) = tan (! 2 ) / tan (! 1 ) tan!! 1 2 +! 4.1. Temperature Effects A significant challenge for our analyses was that different mechanical processes affected the measurements, making it difficult to distinguish the desired variables. For instance, opening an air ventilation door not only causes the temperature in the tunnel to change and generate a pressure wave, it causes the tunnel to vibrate for a significant amount of time. To estimate the effect of temperature variations upon the data, we considered a time scale so that temperature effects dominated the tilt signal. In this case, we chose to use several short (24 hour long) segments. Because long-term trends may still pollute the tilt signal over this time scale, we chose to create a residual (or temperature dominated ) tilt signal by removing the linear trend from each segment and applying a high-pass filter (f corner =10-3 Hz). This corner frequency was chosen because it is much higher frequency than our signals of interest and much lower than the temperature noise. For instance, earth tides have a frequency on the order of 10-5 Hz, while temperature noise tends to have a corner frequency above 10-2 Hz. Figure 2 includes a segment of tilt residuals for a single instrument on the 2000L of the laboratory. Fig. 2. Tilt residuals and temperature measurements over a single-day period. The DC component of tilt measurement is strongly related to temperature.! 2 2 (1) (2) An adaptive least mean squares (LMS) filter was constructed to directly remove temperature effects, but it proved unable to untangle the undesired residual tilt signal from the external desirable mechanical tilt signals. Instead the static temperature correction recommended by Applied Geomechanics was used to diminish the effect of temperature upon the tilt signal. This correction resulted in a static noise level of less than 1 µradian. At this point, any remaining residual tilt measurements are assumed to be noise or transient mechanical signals Groundwater Drawdown Effects Our goal for this section of this analysis was to develop a model of the local aquifer that is capable of predicting the measured tilt signal. We recognized that the hydrology at DUSEL is very complex. For instance, because groundwater is being extracted via an existing shaft and water is fed to the pump by a series of old horizontal drifts, a single vertical wellbore may not adequately describe this system well. Regardless, we chose to develop a simple and analytic model, because it provides significant predictive capabilities, without requiring complex hydrologic models. Because the instruments on the 2000L are located far from active construction, rock excavation will have little effect on our measurements. We can then assume the long-term trends in the tilt data are largely due to the extraction of groundwater. Additional assumptions are that the elastic and hydrologic properties of the aquifer are homogeneous and isotropic, the flow system is radially symmetric, groundwater extraction rates are nearly constant, and recharge from the surficial aquifer is negligible; further, that the tilt rate is proportional to the derivative of vertical effective stress (σ v ) with respect to time and space [3]. For very small tilt rates, a 1-D estimate of tilt was calculated using Eqn. 3, where b is the aquifer thickness, E is Young s modulus, t is time, and r is the distance from the pumping well. The derivative term in Equation 3 was estimated using an analytic model of groundwater drawdown caused by a vertical pumping well [4]. For a radially symmetric, homogeneous, isotropic aquifer, under Darcy flow conditions, the general expression for this term is given by equation 4. W is the well function, and is a function of aquifer geometry, groundwater flow characteristics, and the duration of pumping. Q is the pumping rate, γ is the density of water, K is the bulk hydraulic conductivity, b is the aquifer thickness, E is Young s modulus, and σ is the effective stress. For small time periods, the derivative of the well function is estimated by employing a finite-difference approximation.

4 ( )! arctan % b E! r,t # $ " 2 '! v & "t "r ( ' (3)! 2 '! v!t!r "! Q! 2 W 4! Kb!t!r (4) The modeled effective aquifer thickness was on the order of 1 km, with groundwater being extracted at a rate of about 0.1 m 3 /s using a submersible pump lowered into one of the laboratory shafts. To determine if our methodology was reasonable, a sensitivity analysis was conducted for the case where the rock mass bulk hydraulic conductivity of the rock mass is estimated to be between 10-6 and 10-7 cm/s [5] and the bulk Young s modulus of 80 MPa [6]. The results of the sensitivity analysis are presented in Figure 3. Predicted tilt rates for are highly sensitive to bulk hydraulic conductivity (K) and storativity (S), but given median values of K and S (10-6 cm/s and10-4, respectively) the estimated long-term tilt rate matched the observed tilt rate of 2.4 µradians/day, which is reasonable for this geologic regime [4]. Note that because of the simplicity and non-uniqueness of this estimate, it is difficult to make a more precise estimate of these hydrogeologic parameters. atmospheric data into an adaptive filter capable of removing these effects. This style of filter has been used with moderate success for very long geodetic and atmospheric records (on the order of months) [7] Seismic Direction Analysis Since the instruments were installed in a confined space, and in close proximity to a large iron rail, estimating the orientation of the instruments in space using a magnetic sensor is imperfect. In this instance we applied a methodology using signals external to the laboratory to estimate the orientation of our instruments, in particular, small regional earthquake events and larger teleseismic events. We start by assuming the chosen external signal was elliptically polarized. Since any large teleseismic signal is not likely to be perfectly polarized, a short segment of the time series is selected and band pass filtered. An eigenanalysis of the coherency matrix (see Equation 5) yields information concerning principle direction of energy in the signal [8]. In Eqn. 5, φ is the complex analytical tilt signal, calculated by taking the Hilbert transform of the tilt time series, and * indicates the complex conjugate. For any given segment of tilt, the major eigenvector for this matrix is expected to be oriented parallel to the direction of maximum energy. C = 1 N!! x (i ) $ ' # &!! N "#! y (i ) " * x (i )! * y (i ) $ (5) % i =1 %& Fig. 3. Estimated tilt rates due to the extraction of groundwater form the rock mass. Tilt rate is given as a function of our model parameters: bulk conductivity (K) and storativity (S). The measured tilt rate during January, 2011 is shown in red Atmospheric Effects Atmospheric and surficial loading, especially due to changes in atmospheric pressure and rainfall, can have very significant effects on the measured tilt signals. This loading regime is complex, and non-stationary, and is very challenging to model. We are currently integrating In the absence of any significant site effects, the eigenvectors for a section of the tilt signal should be nearly identical for each instrument; the first arrival of an earthquake (the p-wave) particle motion will be parallel to the direction of propagation. This motion was projected as tilt onto the horizontal plane, with the expectation that there would be a single large eigenvector oriented parallel to this projection, and a much smaller eigenvector resulting from noise and/or imperfect polarization. Since there are a limited number of these events in our database, we focused on determining the relative orientation between the sensors, rather than determining their absolute orientation. Since teleseismic earthquake signals often travel long distances through complex geologic structures, these signals are not necessarily elliptically polarized. Therefore, for a given frequency band, the direction of maximum energy was difficult to estimate with accuracy. We consider the results from many independent events to estimate the orientation of the sensors in order to account for the errors associated with these assumptions.

5 4.5. Earth Tides Our goal is to isolate earth tide signals to obtain an estimate for the elastic properties of the local crust. Earth tides are the result of gravitational forces between the sun, moon and earth. These result in periodic, very low amplitude, tilt signals within the earth. The most significant earth tide constituents have periods of approximately 24, and 12 hours; therefore, a several month long window is required to confidently estimate these signals. A selection of significant earth tide constituents is included Table 1 [9]. We isolated magnitude and phase of the earth tide signals in our data by using a Fourier analysis on very long tilt time measurements for each sensor, after the temperature, groundwater, and atmospheric signals are removed. Although these data are in the frequency domain, we may still apply equations 1 and 2 to infer the magnitude and direction of tilt for each frequency bin. Table 1. Selected earth tide constituents. Name Frequency (µhz) Amplitude (H p /g) O K M S RESULTS 5.1. Raw Data and Tilt Corrections Selected tilt and temperature measurements for the 2000L far station of the laboratory are shown in Figure 4. These data were between January 1 and January 26, There was a sub-linear trend in the data, most likely due to the extraction of groundwater. The overall tilt rate was approximately 3µradians/day, and the orientation of the maximum tilt vector was 54 o counterclockwise from sensor 1. To minimize the departure from the assumed trend, we apply a zero-shift correction of nearly 3µradians/ o C using the lowpass filtered temperature data. This corresponds to the manufacturer s recommended correction values Instrument Orientation A recording of a typical corrected teleseismic tilt signal is included in Figure 5, after the appropriate temperature and groundwater signals are removed. This particular event occurred on March 11, 2011 near Sendai, Japan. The moment magnitude for this event is approximately 9.0 [10]. For this particular event, we chose a segment of the first arrival for use in the eigenanalysis of the covariance matrix. The major eigenvalues for each station were similar, and were more than one order of magnitude greater than the minor eigenvalue. The maximum eigenvectors of the coherency matrix are oriented at 82.7 o and 65.4 o for the near and far stations, respectively. Fig. 4. Measured tilt and temperature data for a single station during January During this period, the average tilt-rate was approximately 3 µradians/day, and the maximum tilt vector is oriented 54 o counterclockwise from sensor 1. These tilt data are corrected for temperature effects, but not for atmospheric loading.

6 Fig. 5. Measured tilt signal following the 9.0 Mw Tohoku, Japan earthquake on March 11, There are very strong P, S, and Rayleigh wave arrival in this record. The duration of measurable shaking in this record is nearly 2 hours Earth Tides The Fourier amplitude spectrum of the tilt data collected during January and February 2011, after the removal of the temperature and groundwater signals, is included in Figure 6. There were two large peaks in the amplitude spectrum that correspond to diurnal (10 to 11 µhz) and semi-diurnal signals (22 to 23 µhz). These regions correspond to many different, and relatively small, earth tide signals; however, the dominant earth tide signals in these regions is likely the K 1 and M 2, respectively. Because there are a limited number of instruments currently recording data in the laboratory, and because of this fairly short period of measurements, we did not attempt to deduce the direction of maximum energy for these signals. Fig. 6. FFT magnitude spectrum for tilt measured over a 26- day long window. The approximate location of the major diurnal (K 1 ) and semi-diurnal (M 2 ) earth tide constituent is indicated. 6. DISCUSSION During our analysis, the base tilt noise level was approximately 0.1 µradians peak to peak, which agreed with our assumption regarding instrument precision. High frequency temperature effects appeared to be negligible for the instruments on the 2000L. Because of differing ventilation characteristics, this may not be the case for tilt-meters to be installed at other locations, so these effects will be considered on a site-to-site basis. The long-term trend in tilt data appeared to be consistent with the simplified model of fluid extraction-driven aquifer drawdown. Over the range of expected bulk conductivities for the rockmass, a storativity between 10-3 and 10-5 predicts the observed 3µradians/day tilt rate (see Figure 3). Furthermore, the direction towards maximum tilt lies roughly towards the main pumping station in the laboratory. Our tiltmeters were capable of measuring global earthquakes with moment magnitudes greater than 6. In addition to the 9.0 Mw 2011 Japan earthquake, we measured significant ground tilt signals from earthquakes in the Chilean and Indonesian subduction zones. We also observed significant tilt signals from smaller, local seismic sources, such as blasting at nearby mines. As would be expected for this style of instrument, the signal is dominated by a strong Rayleigh wave arrival. However, as you can see in Figure 5, our instruments are capable of measuring the major body wave arrivals. Based upon our observations of these different seismic signals, we estimated that our stations had a relative orientation of approximately N 17.3 o. Regardless of the relatively short duration of the tilt records, we were able to observe the effects of diurnal and semi-diurnal earth tide signals in the Fourier amplitude spectrum (see Figure 6). Note the presence of the very low frequency energy below that of the earth tides. We believe this is due to atmospheric effects, and possibly imperfections in the hydrologic model. Due to our limited dataset and the presence of low-frequency energy, we are unable to make estimates concerning the elastic properties of the nearby crust. We do not have the necessary resolution to distinguish the individual tidal signals over these frequency bands; however, as we continue to monitor tilt at various points in the laboratory, we expect to further isolate the individual earth tide constituents.

7 7. FUTURE WORK The Transparent Earth research group is in the process of installing and monitoring new tilt recording stations, and we are developing new recording equipment that is more suitable to the environment at DUSEL. We are developing new, more realistic models for thermomechanical effects, groundwater extraction, and atmospheric loading. A large data bank is being built in order to improve estimates of earth-tides, and increase our confidence in sensor orientations estimates. As time progresses we expect to resolve many of the smaller earth tides that are currently hidden by the largest amplitude constituents. In addition to measuring tilt, we are beginning to integrate a variety of geodetic measurements, such as Interferometric Synthetic Aperture Radar (InSAR) and GPS, to study large-scale processes in the local rock mass. These methods are capable of measuring the deformation of the earth s surface, and have been used to study a variety of processes: tectonic deformations [11], subsidence due to aquifer compaction [12, 13], and mining related subsidence [14]. A major limitation for these studies is that the inversion of these surface data is non-unique. Integrating subsurface tilt measurements will constrain the inversion of these data, providing a more complete understating of largescale deformation processes associated with the development of DUSEL. 8. ACKNOWLEDGEMENTS This research was made possible by grants from the National Science Foundation, CMMI and CMMI REFERENCES 1. Van Beek, J.K., W.M. Roggenthen, M. Magliocco, and S.D Glaser. Rock mechanics and subsurface imaging at DUSEL, Homestake mine. Proc. 3rd CANUS Rock Mechanics Symposium, Toronto, Paper D'Oreye, N Qualification test of a dual-axis bubble-type resistive tiltmeter (AGI-700 series): earth tides recorded and analysed in the underground laboratory of Walferdange. Marees Terrestres Bulletin d'information, 129: Fabian, M., and H.J. Kumpel Poroelasticity: observations of anomalous near surface tilt induced by ground water pumping. J. Hydrology, 281(3): Fetter, C.W Ground-water flow to wells. In Applied Hydrogeology Weinig, W.T., R.S. Popielak, and L.D. Stetler Hydrogeologic conditions at the DUSEL mid-level campus and implications for large cavern design. AGU Fall Meeting 2010, Abstract H13F Respec Consulting & Services Geotechnical engineering summary for the Deep Underground Science and Engineering Laboratory (DUSEL). 7. Braitenberg, C Estimating the hydrologic induced signal in geodetic measurements with predictive filtering methods, Geophys. Res. Lett. 26(6): Jurkevics, A Polarization analysis of threecomponent array data. BSSA. 78(5): Wahr, J Earth tides. In Global Earth Physics: A Handbook of Physical Constants, AGU Ref. Shelf, Vol. 1. ed. T. J. Ahrens, USGS Magnitude 9.0 Near the coast of Honshu, Japan. USGS Earthquake Hazards Program. Retrieved March 15, 2011, from Simons, M., Fialko, Y., and Rivera, L Coseismic deformation from the 1999 Mw 7.1 Hector Mine, California earthquake as infered from InSAR and GPS observations. BSSA, 92(4): Galloway, D., Hudnut, S., Ingebritsen, S., Phillips, S., Peltzer, G., and Rogez, F Detection of aquifer system compaction and land subsidence using interferometric synthetic aperture radar, Antelope Valley, Mojave Desert, California. Water Resources Research, 34: Schmidt, D., and Burgmann, R Timedependent land uplift and subsidence in the Santa Clara valley, California, from a large interferometric synthetic aperture radar data set. J. Geophys. Research, 108(B9): Carnec, C., & Delacourt, C Three years of mining subsidence monitored by SAR interferometery, near Gardanne, France. J. Applied Geophysics, 43,