Processing of GPR data

Raw data taken from GPR are affected by different noises and instability of equipment. The data in this
form, are not suitable for the further analysis. They must undergo a set of transformations in order to
obtain indispensable information.
The processing of GPR data is an immense subject. From a signal processing standpoint, there is a wide
similarity between impulse radar measurement and reflection seismic. Thus there is a whole array of
seismic processing techniques and software available to apply to GPR data.
 Basic Processing Steps of GPR data
 Determination and adjustment of time-zero,

 Basic trace editing (trimming and extraction utilities),

 DC removal

 Dewow and

 Gain manipulation.

Some other utilities –

Modification of the sampling rate (Resample Time-Axis).
Modification of trace spacing – spatial resampling (Resample Scan Axis).
Trace equalization (Equalize).
.
 Determination and adjustment of time-zero

Time-zero adjustment (Adjust Signal Position) - control of the vertical position of the

surface reflection (the place in time where the radar pulse leaves the antenna, and

enters the subsurface is considered as ‘’time zero”). Time-zero correction is necessary

for adjusting all traces to common time-zero position before processing methods. This

point is the time when the first break of air-wave or first negative peak of the trace is

noticed .
Basic trace editing (trimming and extraction utilities)

Radargram size reduction by discarding late-time arrivals reduces the size of the GPR data
matrix by discarding the late arrivals.

DC removal & Wow elimination

It is very important to ensure that the data exhibit a zero-mean property. GPR data include
non-zero mean signals a low-frequency component known as the DC component that
depend on the ground conditions conductive and or wet materials . In general, the DC
component of the GPR signal does not propagate but diffuses into the ground. To correct
for the DC component, a running average filter is applied to the data

This initial direct current (DC) signal component and the very low-frequency signal trend
(or ‘wow’) can generate a distortion of the mean of the A-scan towards values of
amplitude far from zero . This occurrence is partially related to the coupling effect and to
the saturation of the signal by early arrivals
Gain manipulation

 An important aspect of GPR processing is understanding the need for time gain. A major problem with GPR data is that
attenuation of the radar signal in the ground can be highly variable. One can have a low attenuation
environment where exploration depths of tens of meters can be readily achieved. In other situations,
attenuation can be quite high and depths of only 1 or 2 meters can be penetrated.
 Another way of viewing the amplitude of signals versus time is shown by a spherical EM wave spreading into the
ground; it will fall off inversely with distance into the ground, attenuating exponentially due to the
conductivity losses in the sediment. Signals from great depth are very small and may be invisible or
indiscernible in the presence of signals from a shallow depth. Therefore, there is a need to equalize
amplitudes or apply some sort of time gain function that compensates for the rapid fall off in radar signals from deeper
depths.
 The concept of time varying gain is one that applies a gain to the data
which increases with time after the pulse is transmitted. The rise of the gain function is tailored to accommodate the
signal fall-off and signal amplitude versus time.

Two popular gain function applied to GPR data to compensate for the rapid signal fall-off are:
(1) Automatic gain control (AGC), and
(2) Spherical and exponential gain compensation (SEC)
Automatic Gain Control

 With an AGC function, the objective is to equalize the amplitudes all the way down each trace.
 Gains are adjusted by estimating the average signal level at every point on the trace, then computing the gain, which is
inversely proportion to the signal level.
 In areas of weak signal, a gain is large and in areas of strong signal, the gain is small.
 AGC is ideal for monitoring continuity of reflections, but obliterates all amplitude information. Hence, once data has been
processed with AGC, one can no longer make reliable deductions concerning the strength of any particular reflection relative to
other reflections.
 When interpreting GPR data in sedimentary environments, it is important to note that in order to identify high amplitude
reflections a constant gain is required; it is also worth noting that gain amplifies noise and reflections alike.
After dewow After AGC
Inverse Amplitude Decay.

Applies an empirical gain function which exactly compensates the mean or median attenuation observed in a 2-D GPR
section.
The procedure:
(i) Computes the analytic signal for all traces in the GPR section, hence their instantaneous
amplitude.
(ii) Computes a median and a mean amplitude attenuation function, i.e. respectively the median and mean instantaneous
amplitude of all traces in the section.
(iii) Computes empirical best fitting attenuation models with a function of the form

(iv) With N linear parameters and N non-linear parameters.

The gain function is the normalized inverse of the amplitude decay model, i.e. has the form

and produces a practically true white noise series, while preserving relative signal amplitudes
across the scan axis (i.e. at similar 2-way travel times).
 Attenuation characteristics

After gained with inverse amplitude decay

Inverse Power Decay

Data amplification by the Inverse Power Decay applies gain function of the form

g(t) = scale ∗ t^power

 Although application of gain is useful for data display and as an important precursor to some more advanced processing
procedures, it also has the undesirable effect of amplifying various types of ambient and systematic noise (Yilmaz, 1987,
2001). As a result, application of gain should be carefully considered, with the objective of obtaining relevant information
regarding subsurface structure without introducing artifacts.