Detection and remediation of unexploded ordnance (UXO) represents a major challenge. The detection problem is exacerbated by the fact that on sites contaminated with UXO, extensive surface and sub-surface clutter and shrapnel is also present. Traditional methods used for UXO remediation have difficulty distinguishing buried UXO from these anthropic clutter items as well as from naturally occurring magnetic geologic noise, and thus incur prohibitively high false alarm rates. In this research, model-based statistical signal processing techniques are applied to field data from magnetometer and electromagnetic induction (EMI) sensors in order to determine to what degree such an approach results in false alarm mitigation. Features of the target signatures are extracted by inverting the measured sensor data associated with an anomaly using the associated physical, or forward, model. The statistical uncertainty in the feature space is explicitly treated using statistical processors, including generalized likelihood ratio tests and support vector machines, to discriminate targets from clutter. This approach has been evaluated on data collected in a recent field trial that was performed at JPG. Results are presented for one area in which ground truth was known, and for two others in which the ground truth was not known. Substantial reduction of the false alarm rate is achieved for two different platforms, the GEM-3 and the MTADS system. For example, using data from the GEM-3 in one area, the number of false targets was reduced from 181 to 20 with 100% detection of all UXO objects.