Is it a statistically significant anomaly?

Remote sensing, medical imaging, and time-lapse seismic monitoring have common challenges in change detection. In this paper we introduce a statistical procedure developed in medical imaging to determine if anomalies in time-lapse seismic data are statistically significant. Using seismic data redundancy, we construct maps of statistical variation and apply an adaptive thresholding to them. The result is a set of maps of statistical significance of subsurface change. Simulation results show higher sensitivity of the procedure to subtle changes at the level of background fluctuations as compared to conventional data thresholding (based on standard deviations). High noise level and strong data correlation reduce the reliability of the procedure. The procedure applied to real data provides maps that allow quantitative evaluation of interpreted anomalies.