Bayesian Methods Applied to Survey Data From Archeological Magnetometry

Abstract

A model is presented for the interpretation of magnetometer data in terms of archeological features beneath the ground. It describes the detector's response to the assemblage of buried features, incorporating both a spread function and a statistical error process as well as appropriate prior beliefs about the nature of archeological features. The problem is to estimate the magnetic susceptibility of the buried features at each horizontal location. A Monte Carlo Markov chain approach is used to estimate magnetic susceptibilities and all prior parameters. This requires estimation of the normalization constant of the Gibbs prior distribution. The approach is illustrated with both simulated data and measurements from an archeological site. In the latter case, the reconstruction of the buried features corresponds well with the archeologist's observations during subsequent excavation.

KEY WORDS:

• Archeology
• Geophysics
• Image reconstruction
• Inverse data problems
• Metropolis–Hasting algorithm
• Monte Carlo Markov chain
• Normalizing constant