Experimental comparison of repeatability metrics

Pietro (Peter) Giuseppe Gagliardi, Donald C. Lawton

Time-lapse experiments were performed on the nrms repeatability (NRMS), predictability (PRED) and signal to distortion ratio (SDR) repeatability metrics, and the results studied in order to better understand their meaning. First, controlled time-shift, amplitude and additive noise perturbations were made to a baseline seismic trace. Time-shift had approximately linear effects on NRMS of about 15%/ms, subtle hyperbolic effects on PRED and a negligible effect on SDR. Amplitude tests showed that multiplication of the baseline trace by 0.9 resulted in an NRMS value of 10.5% and SDR value of 102.04, while PRED remained unaffected by any amplitude change; analytic equations were found to relate amplitude changes to these metrics. Additive noise experiments revealed that NRMS and PRED are very sensitive to the strength and character of the noise, while SDR seems to be affected little by the noise character.

Second, all three metrics were calculated using a 2D walkaway vertical seismic profile (VSP) dataset from Violet Grove, Alberta, which consisted of three lines. For Lines 1, 2 and 3, NRMS values were 60.6%, 61.4% and 45.2% for horizontal components, and 46.3%, 42.6% and 41.4% for the vertical component. PRED was 0.73, 0.72 and 0.83 for the horizontal components, and 0.82, 0.83 and 0.87 for the vertical component. Finally, SDR was 100.38, 100.29 and 100.70 for the horizontal components and 100.74, 100.85 and 100.79 for the vertical component. The trends of these metrics, while similar, do not always agree with each other, and should be used in tandem to better understand the repeatability metrics of time-lapse data.