Partners in Environmental Technology Technical Symposium & Workshop, Washington, D.C., December 2-4, 2008
SERDP MM1573: Simultaneous Inversion of UXO Parameters and Background ResponseKevin A. Kingdon, Nicolas Lhomme and Stephen D. BillingsSky Research, IncAshland, Oregon
Leonard R. Pasion (P.I.) and Douglas W. OldenburgUniv. of British ColumbiaVancouver, B.C., Canada
Inverting TEM data for Dipole Parameters assuming an Additive Background
MM1573 PROJECT OVERVIEW:Background: Identifying unexploded ordnance (UXO) reliably and efficiently without excavating large numbers of non-UXO is one of the Department of Defense’s most pressing environmental problems. The task of discriminating UXO from non-UXO items is more difficult when sensor data are collected at sites where an electromagnetically active host contributes a large background signal that masks the response of UXO. These include sites contaminated with geological noise originating from magnetic soil and sites requiring UXO detection in conductive sea water. In regions of highly magnetic soil, magnetic and electromagnetic sensors often detect large anomalies that are of geologic rather than metallic origin.
Objective: This project focuses on the accurate recovery of target parameters from geophysical sensor data, even in cases when targets of interest sit in a magnetic or conductive host. Technical objectives include:1. Determining the extent to which a highly conductive or magnetic host interacts with a buried metallic target and 2. Developing improved recovery of target parameters by simultaneously inverting target parameters and the properties of the host material.
AcknowledgementsWe would like to thank Ben Barrowes for collecting MPV data with us in Ashland, Oregon. We would also like to thank Tom Bell and Jim Kingdon for providing us with the Camp Sibert scrap polarizations derived from TEMTADS in-air measurements.
tilt
Investigating the Effect of Magnetic Geology on EMI Data through Numerical Modeling of Maxwell’s Equations
EH3D is a flexible forward modelling program developed at UBC-GIF for calculating the EM fields resulting from a wide range of time domain electromagnetic sources and source waveforms, over a 3D earth that is discretized using a mesh of rectangular cells. These codes were used to model the response of a compact metallic target and a host that has both viscous remnant magnetic as well as conductive properties. Simple geological scenarios were also modelled and compared with multi component data collected over the same geological features.
χ(ω)
Frequency (Hz)
Real
Imag
Real
Imag
Frequency (Hz)
Example: Modelling the VRM Response
EH3D correctly models the viscous remnant magnetization (VRM) response
Target in Halfspace
Target in Freespace
Halfspace Only
σ = 0.1 S/m
σ = 0.1 S/m
σ =1e4 S/m
σ =1e4 S/m
χ(ω)
χ(ω)
Target in Halfspace
Target in Freespace
Real
Real
Imag
Imag
Halfspace Only
σ = 0.1 S/m
σ = 0.1 S/m
σ =1e4 S/m
σ =1e4 S/m
Target in a Conductive HostTarget in a Conductive and VRM Host
EH3D Modelling of a Bump
2. Investigating the effect of terrain on Data• EMI responses generated from irregularities in the topography while surveying.• EH3D was used to model the response for a bump and a trench . • Excellent agreement was observed when comparing the modelling results with data collected using the MPV sensor.
EH3D Modelling of a Trench
Real and imaginary values of the H field for all frequencies modeled in EH3D for a target in a half-space (top 2 panels) and a half-space (middle 2 panels). The bottom 2 panels plot the derived target in free-space solution obtained by differencing the soundings at the center of the loop between the top row of panels and the middle row of panels. The EH3D computed solution for a target in free-space is also plotted in the bottom row panels.
Example 1: Modelling multi-component, multi-static sensor data
• Man Portable Vector (MPV) TEM Sensor Data collected at Sky Research UXO test plot in Ashland, OR
• Soil soundings exhibits the characteristic VRM decay
R2
R3R4
Modelling TEM data collected at sites with magnetic geology Example 2: Modelling EMI Array Data
1. Investigating the Additivity of the Soil and Target responses• The highest conductivity (s=10-1 S/m) half-space was chosen as that is the scenario most likely to produce current channeling which would pose the
most difficulties in the assumption that a target and host soil response can be treated as separate, additive responses. • Soundings are compared for the free-space computed directly from EH3D with a derived half-space achieved by differencing the soundings extracted
at the center of the loop for the EH3D solutions obtained for the target in a half-space and the half-space only models. • The agreement is excellent at all frequencies and additivity is valid for the model considered. Thus a processing procedure that involves subtracting a
background EM response from the data to produce a response that can be modeled as a UXO in free-space is a reasonable procedure.
Magnetic susceptibility model based on lab measurements of Kaho’olawe soil (MM1414)
Orientation Test
• Multi-Sensor Towed Array Detection System (MTADS) EMI data were collected at Camp Sibert, Alabama.• Of the anomalies identified from the MTADS EM61 data, several produced empty holes when excavated.• Due to the presence of magnetic soils we expect that changes in ground clearance (due to wheels moving over
small scale topography) might explain some of the anomalies.• The MTADS sensor does not have an altimeter. Detrended elevation data is used to estimate a ground clearance
Cell 644• An approximately 40 mV anomaly was
detected in the NS lines• The detrended elevation suggests that
there is a variation of approximately 13 cm in the ground clearance
Line 76
Cell 809
Ground Clearance Estimated from Elevation Data
Predicted Geologic Response
MTADS first time channel - Detrended
-600
-400
-200
0
200Original Data
Frequency (Hz)
Sign
al (p
pm)
Tg: RealTg: ImagSoil: RealSoil: Imag
-300-200-100
0100200300
Frequency (Hz)
Sign
al (p
pm)
Corrected Data
Easting (m)
North
ing
(m)
In Phase, 30 Hz
-1 0 1-1
-0.5
0
0.5
1
-700
-600
-500
-400
-300
-200
-100
0
Easting (m)
North
ing
(m)
In Phase, 30 Hz, corrected
-1 0 1-1
-0.5
0
0.5
1
-350
-300
-250
-200
-150
-100
-50
0
Front
RightLeft Center - coaxial
Sensor Tilt (degrees)
Examples of anomalies with soil signals
Example 2. Inversion of Geonics EM63 TEM data at Camp Sibert
Example1. Inversion of Geophex GEM3 FEM data at Camp Sibert
VerticalComponents
RadialComponents
AzimuthalComponents
Rx2,Rx3,Rx4
Observed data Predicted data Residual
Observed data Predicted data Residual
Cha
nnel
1C
hann
el 2
0
Left: Data fit for a pair of soundings. S1 is located away from the target. S2 is located at the anomaly peak. At late times the S2 sounding is dominated by the soil response. Right: Small scale variations in the data due to geology are modelled.
Height Test
Front
RightLeft Center - coaxial
Height Above Ground (m)
We use a Viscous Remanent Magnetization (VRM) model to identify soundings that correspond to soil, and then estimate a spatially smooth background magnetic susceptibility. The magnetically susceptible background is then subtracted from the sensor data, which are then inverted to obtain estimates of the dipole polarization tensor.
Time (ms)
EM
63 R
espo
nse
(mV
)
S1
S2L1
S2
S1
L1
meters EM61 mV Channel 1
Line 15
Line 518
In this example, the detrending processing does not accurately estimate ground clearance. As a result, our estimated ground response appears shifted by 30 mV, but still matches the shape of the measured response.
In this example, we simultaneously estimate the background signal and the target dipole polarizations.
The background susceptibility is assumed to spatially vary as a plane (3 parameters).
Polarizations obtained from TEMTADS in-air measurementsEstimated polarizations when inverting data for 3 unique polarizationsEstimated polarizations when inverting data for 2 unique polarizations
Inversions are carried out in two steps. First the dipole model is used to recover the position, orientation and components of the polarization tensor at all frequencies. Then, the instantaneous amplitudes L(w) at frequency w for the 3-dipole polarizations are fit to the four-parameter model of Miller et al. (2001):
where k is the object amplitude, t is a response time-constant, s is a factor that controls the magnitude of asymptotes at high and low-frequency, and c is a parameter that controls the width of the in-phase peak response.
-346.22
-343.88
-1539.32
-647.1289.98
436.03
-427.48
-127.14Depth = 0.47 m Depth = 0.4 m
Depth = 0.72 m Depth = 1.06 m
Example of soil correction. Soundings that classified as being as soil, are used to form a soil model. A thin plate spline is used to interpolate the soil model to all soundings
Recovered polarization parameters. The time constant parameter separates the 4.2 inch and partial mortars.
Above: The recovered polarizations match the polarizations determined from TEMTADS in-air measurements. The data SNR is not quite large enough for an accurate recovery of the secondary polarizations.
To simplify the calculations we make a number of assumptions:1. The geologic response will primarily be due to viscous remnant magnetization (VRM), and not conductivity. 2. The response for a loop with an arbitrary orientation can be approximated to have the form
This assumption greatly simplifies our calculations, as the functions A and f(t) are function of survey and sensor parameters, and not the geologic properties of the subsurface. 3. Topography does not need to be modeled. 4. The transmitter loop can be approximated as multiple dipole moments.
Imag
Imag
Real
Real
Real
Imag
Imag
Imag
Real
Real
Raw Elevation Detrended Elevation Observed vs. Predicted