HIGH MARKS AT JPG PHASE III FOR NAEVA GEOPHYSICS

Results from Phase III of the Unexploded Ordnance Advanced Technology Demonstrations at Jefferson Proving Ground (JPG) show that ordnance detection statistics have improved markedly over Phase II, with some demonstrators using magnetometer and electromagnetic induction combinations yielding probability of detection (P_D) rates over 90%. Sponsored by the U.S. Army Environmental Center and the Naval Explosive Ordnance Disposal Technology Division, the demonstrations help assess technologies suitable for the detection, identification, and excavation of unexploded ordnance. Selected from 32 applicants, the 15 Phase III participants performed their demonstrations within preset parameters, or scenarios. Among the best performers, NAEVA Geophysics Inc., surveyed two 10-acre tracts, the Aerial Gunnery Range and the Artillery and Mortar Range (scenarios 1 and 2, respectively).

Instrumentation

NAEVA used Geonics EM-61 electromagnetic units and Scintrex Smartmag SM-4 cesium-vapor magnetometers. Bot the EM-61 and SM-4 systems are man-portable. Together those instruments can detect ferrous and nonferrous metallic objects. The EM-61 metal detector is capable of sensing small, shallow metallic objects (ferrous or nonferrous), while the magnetometer is preferred for detecting large, deeper ferromagnetic objects beyond the range of reliable EM-61 metal detection.

The Geonics EM-61 is a time-domain electromagnetic instrument consisting of two air-cored coils, batteries, processing electronics, and a digital data recorder. The coils are arranged so that the larger coil (EM source and receiver) lies 40 centimeters below a second receiver coil. Secondary voltages induced in both coils are measured in millivolts. The EM-61 is designed to detect shallow metallic objects (ferrous or nonferrous) with good spatial resolution.

The Scintrex Smartmag SM-4 cesium-vapor magnetometer measures the total magnetic field with a sensitivity of ` 0.01 nT (range 15,000 to 100,000 nT) at rates from 1 to 10 samples per second. The SM-4 system includes the cesium sensor, associated electronics, a carrying harness, ENVI control console, ENVIMAP operating software, and rechargeable batteries.

Data Acquisition

EM-61 data were collected over the Aerial Gunnery and Artillery and Mortar Range with a sensor separation of about 1 meter. Initially, three EM-61 units were "ganged" together providing a 10-foot sweep width. The EM-61’s operated in wheel odometer mode, collecting data every 0.63 feet along lines. Even with pulse synchronization, some increased background noise between instruments was experienced in the "ganged" mode. Consequently, individual EM-61 units were separated and operated independently. Magnetic data were collected in continuous reading mode (10 readings/second) with a 5-foot line separation. A separate base station magnetometer acquired data for magnetic diurnal drift corrections.

With permanent grid nodes already established on the demonstration sites on a 100-foot by 100-foot basis, the survey lines were controlled by chain on 10-foot-spaced north-south survey grid lines with references points every 20 feet in order to assure better than ` 1 foot accuracy.

Data Analysis and Interpretation

The EM-61 data were plotted in profile for each survey line and interpreted using Geonics DAT-61 software. The magnetic data were processed and analyzed using the MAGFIT method. MAGFIT is a unique and proprietary computer program developed by G. Hunter Ware and Hunter Andreas Ware. It scans the theoretical anomalies of a very large number of magnetic dipole models (all locations, depths, orientation, and dipole moments of interest) over the field data, and identifies the best models using a "best least squares fit" criteria. MAGFIT yields model fit contours (in percentages) around the best fit location, in plan or cross section. It also calculates estimated mass and depth for the buried ferromagnetic objects. Location, depth, and size estimates from both the DAT-61 and MAGFIT methods were integrated before predicting the final target locations and depths.

Survey Results

NAEVA’s survey results for the combined scenarios yielded a P_D of 94% and a false alarm ratio of 1.96 (that is, the number of false alarms per ordnance item detected). That P_D rate ranks the highest among those geophysical demonstrators who participated in at least two search scenarios. Based on the false alarm ratio, which is a measure of the likelihood of ordnance in target excavations, over one-third of NAEVA’s target excavations would have resulted in ordnance finds. That result is better than other combined-scenario demonstrators with similar detection rates. Finally, NAEVA’s false alarm rate, 24.8 false alarm per hectare, is relatively low among all demonstrators.

Summary Comments

Regarding instrumentation, both the Geonics EM-61 and Scintrex SM-4 Smartmag units are high-quality commercially available instruments. Indeed, the Geonics EM-61 are perhaps the most frequently used instruments by JPG demonstrators. Consequently, our success should not be attributed to any "newly developed secret instrumentation" that is more "sensitive" than those used by competitor demonstrators. We are of the opinion that although the quest for new and innovative instruments should continue, the sensors we utilized at JPG already typically have an instrument noise level well below that of the terrain noise. Navigation and grid control were accurately maintained by chain rather than by using more "sophisticated" but oftentimes less accurate GPS methodology.

Regarding data processing and interpretation, the DAT-61 software is commercially available, and while the MAGFIT software is proprietary, its theoretical basis has been published.

We believe our success is attributed first to our experienced field crews carefully executing precise field procedures during data collection, assuring the first step in a successful program: high quality raw data. Second, regarding processing and interpretation, we did not rely solely on computer software to make our target selection. Data were processed and analyzed using model-based software, but all final target selection/discrimination was reviewed manually, so that human judgement could be applied. Our results demonstrate the validity of such interpretational procedures, providing the data collection locations are accurate, and the signal (message) to noise (instrument plus terrain) ratio of the instruments is satisfactory. Further efforts in research and development need next be applied to the task of enhanced target characterization.