Using Fracture Seismic methods to map fluid-conducting fracture zones makes it important to understand fracture connectivity over distances greater 10–20 m in the Earth’s upper crust. The principles required for this understanding are developed here from the observations that (1) the spatial variations in crustal porosity are commonly associated with spatial variations in the magnitude of the natural logarithm of crustal permeability, and (2) many parameters, including permeability have a scale-invariant power law distribution in the crust. The first observation means that crustal permeability has a lognormal distribution that can be described as κ ≈ κ 0 exp ( α ( φ − φ 0 ) ) , where α is the ratio of the standard deviation of ln permeability from its mean to the standard deviation of porosity from its mean. The scale invariance of permeability indicates that αϕο = 3 to 4 and that the natural log of permeability has a 1/k pink noise spatial distribution. Combined, these conclusions mean that channelized flow in the upper crust is expected as the distance traversed by flow increases. Locating the most permeable channels using Seismic Fracture methods, while filling in the less permeable parts of the modeled volume with the correct pink noise spatial distribution of permeability, will produce much more realistic models of subsurface flow.
Sonic velocity and electrical resistivity logs run to a depth of 3.5 km in crystalline rock near the San Andreas fault at Cajon Pass in southern California correlate over scale-lengths both small (sub-metre) and large (tens to hundreds of metres). No such correlations are seen with the more lithologically sensitive natural gamma intensity log. The correlation between the sonic velocity and electrical resistivity logs suggests that a non-lithologic property of the crystalline rock controls fluctuations. In situ fracture intensity is a logical candidate for the controlling rock property. The fluctuations of the individual sonic velocity and electrical resistivity logs are examined with the Hurst rescaled range parameter over borehole log intervals 1.5 m < L < 1500 m. For log fluctuations arising from a scale-invariant physical process the Hurst rescaled range scales with data interval as LH, 0 < H < 1. A purely random sequence of in situ fractures produces a scaling exponent H= 0.50. Fluctuations in the Cajon Pass sonic velocity and electrical resistivity logs yield H~ 0.70 evidence that in situ fracture sets tend to occur in clusters rather than at purely random intervals. The tendency for fracture clustering over log intervals 1.5 m < L < 1500 m suggests that fracture formation is a fractal process independent of length-scale in which larger fracture intervals form from clustering of numerous smaller fracture intervals. Seismic reflectivity derived from the borehole sonic velocity log is also scale independent over the range of data intervals 1.5m < L < 1500 m with a Hurst exponent H= 0.21. If we associate fracture clustering with crustal fault formation, the Cajon Pass borehole sonic velocity and electrical resistivity logs predict that crustal faults scale fractally with fractal dimension D̃ 2.30. The equivalent b-value for earthquake size distribution is b~D/2~ 1.15. On this hypothesis observed b-values < 1.15 indicate a tendency for earthquakes to cluster on existing (weak?) faults. A scale-independent mechanics for crystalline rock fracture formation in which larger scale fractures form as clusters of smaller scale fractures provides a mechanical link between the small pervasive stress aligned flaws, cracks and microfractures which can impart anisotropy to crystalline rock and the larger scale fractures associated with finite strain and faulting. Thus in crustal regions of active but low strain, fracture anisotropy is observed to be aligned with the inferred maximum principal stress, while in actively faulted crust, fracture anisotropy is observed to be more nearly fault parallel as if the fractures are aligned by the finite strain faulting process.
Theory, laboratory and field evidence reported elsewhere demonstrate that shear-wave splitting monitors the low-level pre-fracturing deformation of the crustal rock which is driven by the response of fluids in cracked rock.
Five magnetotelluric stations were taken in the Yellowstone region. Sufficient coherent signal was obtained in the frequency band 10 −4 to 10 −1 hz to yield useful apparent resistivity (ρ a ) spectrums. At stations near Yellowstone the resistivity was linearly proportional to ω at long periods; this is due to the existence of a high conductivity substrate at some depth. The effective thickness of the low conductivity surface zone as inferred from the intercept of the response curve is consonant with the geological setting; it is thin (<5 km) in the Yellowstone thermal area, and thicker on the Snake River Plain (≈ 15 km). Stations well away from Yellowstone show ρ a ∼ ω 0.7 , suggesting a lower, gradient of conductivity in the substrate. Precise inversion of the data is not practical: however, the comparison of ρ a at neighboring stations shows distinct differences in electrical response of the crust which may be discussed in a geological context.
The sensitivity to oil-water fluid-substitution events of Uniwell sourcing and sensing seismic waves in the same borehole is degraded by tubewave noise generated by the source. To assess some abatement measures, a ten-level clamped vector-motion sensor string was ganged with a high-frequency or a low-frequency seismic source to record data on tubewave noise levels.
Water levels in the Los Angeles Aqueduct in southern California fluctuate in a manner that are not easily attributable to normal aqueduct operations. Simple hydraulics suggests that large scale earth tilt can register as water level anomalies with a sensitivity of about .01 ft/microradian. Two aqueduct anomalies which coincide spatially and temporally with independently observed deformational phenomena are used to explore this suggestion.
The downhole orbital vibrator (DOV) acts as a rotating acoustic point force coupled to the borehole fluid. Operated as a Vibroseis source modulated over rotational frequencies 50-350 Hz, the DOV can generate crosswell seismic signals at well separations up to 800 m in oilfield sediments. Source-wavelet stability and crosswell seismic traveltime resolution are estimated from an ensemble of 13 000 wavelets recorded at a crosswell seismic facility with 300 m source-sensor separation. DOV auto-correlation source wavelets Si cross-correlated against the ensemble mean wavelet Smn give correlation coefficients γi=Si★Smn having mean γmn≈ 99.97 per cent and standard deviation γrms≈ 0.03 per cent. S wavelets give observed traveltimes τi with normalized standard deviation Δτrms/τmn≈ 0.03 per cent. This level of seismic monitoring source stability and crosswell traveltime resolution offers considerable promise for accurate, cost-effective time-lapse seismic imaging of active geofluid reservoirs. Application of stable DOV waveform production to time-lapse seismic imaging is, however, affected by the dual-wavelet nature of rotary motion cross-correlations. In general, DOV cross-correlation signals mix time-symmetric (even) wavelets χcc(t) ≅ cos(ω(t)t) ★ cos(ω(t)t) ≅χcc(−t) and time-antisymmetric (odd) wavelets χcs(t) ≅ cos(ω(t)t) ★ sin(ω(t)t) ≅−χcs(−t), where ★ denotes correlation, ω(t) describes the modulation of rotational frequency, and time-reversal t↔−t equates to interchanging clockwise/counter-clockwise (cw/ccw) DOV rotations. Cross-correlating sensor motion along source-sensor axis x with source motion along axis g mixes wavelet symmetries as χg(t) ≅ cos φχcc(t) + sin φχcs(t) where cos φ=g·x and angle φ orients the DOV relative to the source-sensor axis. DOV orientation uncertainty Δφ affects the sensor wavelet apparent traveltime as Δτ≈ 1.3(Δφ/2π)Tmin, Tmin the period of peak modulation frequency. Since source orientation uncertainty can generate apparent traveltime uncertainty as large as 1-2 ms, time-lapse seismics cannot effectively ignore DOV orientation. Traveltime resolution can be improved to order 0.1 ms without knowing DOV orientation if traveltimes are computed using even/odd wavelets composed by summing/differencing cw/ccw wavelets. Prospects for time-lapse resolution improve if observers can orient the DOV. A DOV equipped with a collar of rotation-phase point detectors surrounding the rotating mass permits observer selection of the monitor sensor, hence controlling effective source orientation. DOV orientation can be guided by an on-board biaxial tiltmeter (for borehole tilt δ along an axis making angle φ with DOV sensor orientation, biaxial tilts T1 and T2 fix φ as T1=−cos φtan δ and T2= sin φtan δ). Numerical simulation of full-sensitivity time-lapse reservoir monitoring suggests it is feasible to resolve migrating oil/water substitution volumes of characteristic dimension 20 m with crosswell transmission data over ≈600-800 m offsets, or in backscatter data at offsets ≈100-200 m using a DOV and sensor array operating in a single well.
The nature of geological heterogeneity is well understood in proximity to a wellbore but is sparsely sampled laterally. Heterogeneity is both highly unpredictable and difficult to map with surface reflection seismics, but high productivity trends in tight gas reservoirs can be efficiently identified with precision crosswell seismic time-lapse travel-time data.
Crosswell seismic data recorded at 620–650 m offsets in an oil-bearing sand/shale reservoir formation at the Liaohe Oil Field, northeast China, provide robust evidence for waveguide action by low-velocity reservoir layers. Crosswell-section velocity models derived from survey-well sonic logs and further constrained by observed waveguide seismic wavegroup amplitudes and phases yield plausible evidence for interwell reservoir–sand continuity and discontinuity. A pair of back-to-back Liaohe crosswell vector-seismic surveys were conducted using a source well between two sensor wells at 650 and 620 m offsets along a 200-m-thick reservoir formation dipping 7° down-to-east between depths of 2.5 and 3 km. A downhole orbital vibrator generated seismic correlation wavelets with frequency range 50–350 Hz and signal/noise ratio up to 5:1 over local downhole ambient noise. The sensor wells were instrumented with a mobile 12- to 16-level string of clamped vector-motion sensor modules at 5 m intervals. Using 5 m source depth increments, crosswell Surveys 1 and 2 cover source/sensor well intervals above and through the reservoir of, respectively, 600 m/600 m (13 000 vector traces in 9 common sensor fans) and 300 m/560 m (7000 vector traces in 7 common sensor fans). Survey 1 common sensor gathers show clear, consistent high-amplitude 20 ms waveletgroup lags behind the first-arrival traveltime envelope. Such arrivals are diagnostic of seismic low-velocity waveguides connecting the source and sensor wells. Observed Survey 1 retarded wavegroup depths tally with source and sensor depths in low-velocity layers identified in sonic well logs. Finite-difference acoustic model wavefields computed for waveguide acoustic layers constrained by well-log sonic velocity data match the observed waveguide traveltime and amplitude systematics. Model waveforms duplicate the observed m-scale and ms-scale sensitivity of waveguide spatio-temporal energy localization. Survey 2 crosswell data, in contrast, provide no comparable evidence for waveguide action despite a sensor-well sonic well log similar to that of Survey 1. Instead, acoustic wavefield modelling of Survey 2 data clearly favours an interpreted waveguide model with 10° downdip interrupted by a 75–100 m throw down-fault near the sensor well. The absence of clear waveguide arrivals is adequately explained by dispersal of waveguide energy at the fault discontinuity. Auxiliary well sonic velocity and lithologic logs confirm the model-implied 75–100 m of down-throw faulting near the sensor well.
