The presence of unexploded ordnance (UXO) in coastal regions and at sites designated for base realignment and closure (BRAC) poses a severe risk that must be addressed by the Department of Defense before sites can be turned over to the public. Presently, no effective capability exists to survey underwater areas successfully, map the locations of UXO, and distinguish them from clutter. Although advances have been made with magnetic and electromagnetic systems, they are limited in their effectiveness either because they provide low resolution or their depth penetration in saltwater environments is limited. In comparison, current sonar methods offer an improved range of penetration in saltwater, but they are limited in their ability to detect UXO and clutter buried in seafloor sediments. To reduce the significant costs associated with base cleanup efforts, an urgent need exists for reliable methods that work in marine environments and are capable of penetrating the seafloor sediments.

The objective of this limited-scope project was to investigate whether narrow-band resonance-scattered waves could be observed in a small data set collected by a sonar platform in a controlled pond environment and whether these waves could be used to locate UXO in the pond sediments. Furthermore, researchers investigated whether UXO could be characterized by their size and filler velocities and whether medium velocities could be derived from the field data.

Technical Approach

Sonar data of proud UXO were acquired by the Applied Physics Laboratory, University of Washington (APL-UW), at the pond facility of the Naval Surface Warfare Center, Panama City (NSWC-PC), Florida. Researchers investigated whether resonance scattered waves could be observed in this data set and whether these waves could be used to locate UXO in the pond sediments. Numerical modeling was employed to demonstrate the nature of resonance scattering and how resonance waves can dominate the scattered wavefield of proud and buried UXO-like objects.


This project found that resonance scattered waves were present in all numerical data and in the cases of a proud sphere and a buried cylinder. In all other cases of proud UXO, the field data revealed frequency-dependent scattered waves that were part of the direct reflected waves. Imaging UXO with monochromatic data produces surprisingly good results considering that in most investigated cases a small fraction of the data with very low amplitudes were used in comparison to data typically used for Kirchhoff migration or Synthetic Aperture Sonar (SAS) imaging. The images derived from resonance scattered waves were generated by data obtained from narrow passbands throughout the whole frequency spectrum such that different parts of the UXO can be illuminated separately. This has the advantage of imaging, for example, a cone-shaped tip of a UXO separately from its broader tail, which helps to interpret the resulting images. The fact that the resonance images were obtained from much smaller signal amplitudes suggests that monochromatic source signals with longer time duration (i.e., comparable to typically used broadband pulse times) would greatly increase the signal-to-noise ratio. Consequently it would be possible to image smaller objects in sediments that would not be visible when illuminated by a broadband signal. Range resolution is reduced when imaging is done with monochromatic signals. However, the image resolution can be dramatically improved by adding data from different azimuth as was demonstrated for the cases of a proud sphere and bomb. It is noted, however, that the images obtained with resonance migration are intended for detection and localization only, while characterization of the object is based on spectral properties.

The resonance waves were manifested by a range of spectral peaks in the amplitude spectrum. The complex spectral pattern depends on the shape, the dimensions, and the internal properties of the UXO and the properties of the embedding medium. Therefore, the spectrum can be used as a spectral fingerprint to characterize the object after it has been detected and localized in space. To achieve this goal, a database with typical UXO-shaped objects located in sediments needs to be compiled, to include the effect of intrinsic attenuation that cannot be easily modeled with the current knowledge and computational resources. Results from spectral analysis of numerical data from a single UXO under different orientations suggest that the spectral peaks associated with the different spectra were stable and could be matched at low and high frequencies. The outcome suggests that the resonance peaks can be used to characterize the objects when matched to a spectrum from a pre-computed database. In contrast, two cylinders with small differences in size produced resonance peaks that were not comparable, such that the two objects could be distinct by their spectral character. In a previous study, the minimum difference in typical filler velocities was determined to be 150 m/s. A numerical experiment was conducted to determine whether the smallest difference in typical filler velocities of 150 m/s is detectable in the frequency spectrum. The results for identical cylinders with a difference in P- and S-wave velocity of 150 m/s and 87 m/s, respectively, produced relative shifts of 2.5% in the resonance peaks, suggesting that the peaks could be used to distinguish UXO filler.

Velocity estimates of the background medium have been performed based on the fidelity of the resonance images. A range of velocities including the correct value will produce a suite of images that are progressively defocused the more the migration velocity differed from the correct value. Therefore, the sharpest migration image will yield the correct medium velocity. This approach is independent of the number of objects imaged, because the principle of focusing and defocusing is universal.


As this technique is developed, it could be applied in field situations to locate and characterize buried UXO in reacquisition missions using current sonar technology. Data processing and estimation of the properties from the resonance spectra to locate and characterize buried UXO are fairly robust, such that the process can be automated in field applications, facilitating ease and speed of operation.