Specifications

University of Pretoria etd – Combrinck, M (2006)
Sounding 1 (Figure 4-11) would be considered “noise-free” in that none of the data points
would probably be masked (ignored) in manual interpretation of the data. However,
comparing the smoothed and unsmoothed results it is clear that it does contain some noise
which can be filtered out successfully with the weighted moving average filter. The
unequally spaced points are less sensitive to the noise and after smoothing there is no
noticeable difference between the three different methods of differentiation. (With field
data it is not possible to calculate percentage errors and smoothness of conductivity-depth
curves are used as an indication of accuracy based on the smoothly dissipative nature of
TEM currents.)
(c)
1.E-03
1.E-02
1.E-01
1.E+00
1.E+01
1.E+02
1.E+03
0 100 200 300 400
Depth [m]
Imaged Conductivity [S/m]
(d)
1.E-03
1.E-02
1.E-01
1.E+00
1.E+01
1.E+02
1.E+03
0 100 200 300 400
Depth [m]
Imaged Conductivity [S/m]
(a)
1.E-03
1.E-02
1.E-01
1.E+00
1.E+01
1.E+02
1.E+03
0 100 200 300 400
Depth [m]
Imaged Conductivity [S/m]
Lagrange Three-Point
Cubic Spline
In ve rse of Inte grati o n
(b)
1.E-03
1.E-02
1.E-01
1.E+00
1.E+01
1.E+02
1.E+03
0 100 200 300 400
Depth [m]
Imaged Conductivity [S/m]
Figure 4-12: S-layer differential transform results for Sounding 2.
Sounding 2 contains slightly more noise (Figure 4-12). The most noticeable effect is that
the re-sampling to equally spaced data points causes the method to fail. (This is mostly due
to re-sampling of the cumulative conductance (S) curve which is differentiated in step 5 of
the algorithm. This curve, although smooth, sometimes forms a relation rather than a
52