Accuracy and Precision of GPR Velocity Models Obtained From Semblance Analysis of CMP Gathers

HR: 14:20h
AN: NS33A-02 [Abstracts]
TI: Accuracy and Precision of GPR Velocity Models Obtained From Semblance Analysis of CMP Gathers
AU: * Booth, A
EM: a.d.booth99@leeds.ac.uk
AF: School of Geography, Faculty of Earth and Environment, University of Leeds, Woodhouse Lane, Leeds, LS2 9JT, United Kingdom
AU: Clark, R
EM: r.clark@see.leeds.ac.uk
AF: School of Earth Sciences, Faculty of Earth and Environment, University of Leeds, Woodhouse Lane, Leeds, LS2 9JT, United Kingdom
AU: Murray, T
EM: t.murray@swansea.ac.uk
AF: Department of Geography, School of the Environment and Society, Swansea University, Singleton Park, Swansea, SA2 8PP, United Kingdom
AB: Interest is growing in the use of ground penetrating radar (GPR) methods for quantifying subsurface properties (e.g. porosity, water content) derived from EM wave interval velocity, v_INT. This velocity is usually calculated from Dix's Equation with velocity and time picks obtained from semblance analysis of reflection moveout times in common midpoint (CMP) data. However, this process leads to imprecision and inaccuracy in velocity estimates (and resulting estimates of subsurface properties) for 3 reasons. (1) The CMP geometry (through the range of offsets and thus moveout times) controls observational error and hence imprecision in `root-mean-square velocity', v_RMS. (2) Dix's Equation only delivers true v_INT when input velocities are v_RMS. The latter are not recovered from actual, non-hyperbolic, moveout times; instead they define `stacking velocities', v_ST, which in turn yield a systematically inaccurate `interval stacking velocity', v_IS. (3) Peak semblance response is to the maximum amplitude of the GPR wavelet, typically the second or third half-cycle, rather than first-break travel-times. This delay from first-break results in a further systematic bias of v_ST to slower values. GPR velocity analyses rarely recognise these issues; this paper evaluates the severity of the problem, and suggests field procedures and remedial measures to minimise its consequences. Analyses of synthetic data show that v_ST uncertainty is <1% if a reflection exhibits a ratio of [moveout/wavelet period] of >7.5. A CMP acquisition must contain traces to sufficiently far offset to fulfil this criterion. Such a v_ST is then used to predict the v_ST that would be recovered from near- offset traces, which in turn more accurately matches v_RMS. Finally, in order to simulate first-break times, synthetic travel-times generated from near-offset v_ST are subtracted from the (assumed constant) delay between the wavelet first-break and its maximum amplitude. For 50 MHz mixed-phase energy (delay = 9.4 ns) reflected at t₀ = 200 ns in an overburden velocity of 0.1 m/ns, simulation of first-break travel-times reduces the discrepancy between v_ST and v_RMS from -16.0% to +0.5%. The initial error in v_RMS reduces as the delay to maximum amplitude becomes a smaller fraction of t₀ (e.g. shorter wavelet period and/or longer t₀). For a glaciological field example, this approach modifies a water content estimate from 1.71% to 1.18% (-0.38, +0.15), due to improved accuracy and precision of v_INT.
DE: 0720 Glaciers
DE: 0910 Data processing
DE: 0925 Magnetic and electrical methods (5109)
DE: 1835 Hydrogeophysics
SC: Near-Surface Geophysics [NS]
MN: 2007 Joint Assembly