The intent of the project was to address unexploded ordnance (UXO) detection in complex and hostile soil environments such as Kaho'olawe Island, Hawaii and the Waikoloa Maneuver Area, Hawaii. Geophysical target signatures recorded from electromagnetic induction (EMI) sensors, such as the Geonics EM61, contain extreme levels of background noise caused in large part by magnetically permeable soils. The spatial variation of the soil effects is not separate and distinct from the signatures of buried UXO at a single frequency or time gate. This minimizes the effectiveness of spatial filtering for separating UXO signatures from soil effects. Existing fixed-geometry, transmitter/receiver coincident EMI systems have not produced data sufficient to reliably detect UXO in the presence of strong geologic effects.

Technical Approach

The researchers’ concept was that the viscous remanent magnetism (VRM) soil responses could be addressed by use of differential measurements. That is, measurements that excite a different response in the soil compared to metallic items. Four types of differential measurement were considered:

  1. Receiver height: Intended to exploit the different fall-off of the soil response compared to that of a compact metallic object;
  2. Transmitter field direction: The horizontal field of a vertical dipole transmitter is null-coupled to a soil with laterally homogeneous soil properties. Thus, the soil should exhibit a weak to non-existent horizontal field in contrast to a compact metallic object;
  3. Transmitter loop size: Original concept used to motivate this study, but subsequently it was discovered that the soil and metallic responses scale in the same way with loop-size; and
  4. Transmitter waveform: Intended to exploit different responses in metal targets compared to soil through varying the charge-up time.


The first two differential measurements listed above assume that the soil can be approximated by a homogeneous half-space. At the site visited on Kaho'olawe Island, there were significant variations in the concentration of magnetic materials over short length scales. In addition, micro-topography caused large amplitude, short wavelength anomalies in the EMI data, due both to changing sensor-ground geometry and variations in sensor orientation. Thus, the soil response would not be well modeled by a half-space and the researchers’ infer that the first two differential measurements listed above would have met with little success.

The researchers postulated that micro-topography could have been the cause of a large number of the 30% of anomalies excavated on Kaho'olawe that were attributed to geology.

This work concentrated on differential measurements based on variations in transmitter waveform. The main reason for this was that the time-domain electromagnetic (TEM) response of a soil is dominated by the VRM (and generally not the soil conductivity) and that the form of the TEM decay from VRM material is independent of its spatial distribution. Thus, while the amplitude will change with micro-topography, coil orientation and lateral distribution of VRM, the form of the TEM decay is invariant. This makes any technique based on this phenomenon attractive from a practical point of view.

Early on in this project, soil compensation methods were considered that involved estimating the soil response from a single transmitter waveform. These types of methods can work well, but there is the possibility that a UXO will fit the soil model, and hence not be flagged for excavation. Such an example was presented from Kaho'olawe involving a 5in high explosive (HE) round. This provided strong incentive to develop a differential measurement methodology.

On Kaho'olawe Island and at the Waikoloa Maneuver Area, field studies were conducted to evaluate the effectiveness of differential measurements based on varying transmitter charge times. EM-61MKII measurements with three different transmitter on-times of 10ms, 4ms and 2ms were collected. Analysis showed that for typical UXO targets, differential measurements based on these varying charge times were no more effective at suppressing the soil response than soil compensation methods based on a single charge time. The reason for this is that the time-constants of most UXO are long compared to the latest time channel measured by the EM-61. This means that both soil and metallic objects exhibit similar differential responses. If the time constants of UXO were significantly smaller than the latest time channel, these differential measurements would have excited a different response compared to the soil.

The differential technique based on varying the charge time would only work for typical UXO if the measurement range extended past the object's characteristic time-constant. A 5cm diameter steel sphere, as a point of comparison, has a time constant of 40ms. This would represent a UXO like-object of the small to medium range. Larger UXO would have even larger time constants (the particular time constant excited depends on the UXO orientation). 40ms represents a long time after pulse-shutoff; for example, the EM-61 measures out to around 3.5ms, while the EM-63 extends that range to 25ms. Thus, existing commercial sensors stop measuring before the differential effect would likely be observed. In addition, even if the sensors did measure that far in time, the signal-to-noise ratio would be small. Thus, it appears that differential measurements based on variable charge time are unlikely to improve UXO detection performance in magnetic soils.


While the differential measurements trialed in this project did not result in an appreciable improvement in soil compensation techniques, the study did improve the UXO community's understanding of magnetic soils. Prior to this study, preliminary work on the importance of the frequency dependence of susceptibility (i.e. VRM) on the soil response had been conducted. A good understanding of the physical mechanisms that control the effect of soil on a metal detector was obtained during this study. In particular, this work now has the ability to predict the soil response from an arbitrary transmitter waveform. These predictions depend on the assumed model of magnetic viscosity (i.e. that it is caused by magnetic domains with a log-uniform distribution of time-constants). Several studies are ongoing that involve the development of instrumentation to measure the susceptibility at multiple frequencies. This will confirm the assumption that soils can be modeled with a log-uniform distribution.