Electromagnetic induction in subsurface metal targets: Cluster analysis using local point-pattern spatial statistics

The detection and localization of statistically significant spatial clustering of buried metal artifacts for applications, including archaeological, environmental site assessment, and civil infrastructure mapping, can be performed reliably, inexpensively, and noninvasively using transient electromagnetic (EM) geophysics. The local G(i)* (d) statistic based on a weighted spatial autocorrelation were used to detect and localize statistically significant clustering of buried metal objects from gridded electromagnetic responses acquired at a seeded control site and at an active historical archaeological site in Paint Rock, Texas. This statistic compares the local EM millivolt response values about point i radially to the global mean and variance to calculate the G(i)* statistic as a standard variate at various length scales, which can therefore be assessed for significance. Because the local G(i)*(d) statistic compares local to global averages, it spatially,compacts the amplitude response signal of the EM data. The resulting spatial cluster maps conformed well to the known distribution of metal targets buried at the control site, and enabled a successful excavation strategy with a significant reduction in false positives encountered at the archaeological site, when compared with more traditional Methods.