March 2021
Dr. Jan Francke
jfrancke@[Link]
GPR Processing Basics
Historically, the vast majority of GPR instruments which were used commercially output data
which were processed, interpreted, and presented in a report format, regardless of the application.
As GPR manufacturers sought to widen acceptance of the technology and appeal to non-
geophysicists, the concept of site mark-outs, whereby the GPR is used to mark-out targets in the
field (usually with spray paint) became commonplace. Indeed, the acquisition software and
dataloggers for the GPR systems now contain sufficient processing and interpretation power to
allow an on-site evaluation of both 2D and 3D data.
There may presently be a shift back to reporting requirements for civil infrastructure applications
of GPR, necessitating some post-processing of the acquired data.
There are numerous processing options available to the user and some equipment manufacturers
(e.g. GSSI) store recorded data in a pre-processed form. The vast majority of civil infrastructure
surveys can be addressed using a handful of processing steps, which are listed in the
recommended order below.
Time Zero Correction
The radar measures the two-way travel time between the time of transmission to the time of
reception. The receiver generally starts recording at the instant the transmitter is triggered. As
some time is required for the signal to travel through the electronics, the antenna, the air (if using
an air launched antenna) etc, there is normally a zone of data above the “direct arrival”, which is
defined at the arrival time of the energy travelling directly through the air from the Tx to the Rx.
This zone, above “time zero”, must be removed before any further processing can be performed.
The time offset of this direct arrival on modern GPR systems is constant and a single time zero
correction can be applied for the entire profile. However, on some particularly older GPR systems,
the time offset for the direct arrival can drift and the correction should be done on a trace-by-trace
basis.
1
Page
Figure 1 – Data recorded pre-“time zero”
Figure 2 – Data after time-zero correction
Gain
GPR energy is attenuated the moment it is emitted by the transmitter. Within the electronics there
are mismatches in impedance which cause small losses, but the largest losses come from Ohmic
losses (conductive soils like clays), geometric losses (the GPR is like a torch, not a laser), and
scattering losses (gravels and cobbles scatter energy in all directions). Applying a gain function
is an attempt to account for these losses.
There are many approaches to loss compensation (gains), with the most common being AGC
(amplitude gain correction) and some form of energy decay compensation. All to often,
inexperienced users rely on AGC as it appears to enhance interpretability more than other gains.
Although this is generally true, AGC gains cause all relative amplitude information to be lost since
an AGC gain attempts to equalise all amplitudes. Thus, a metal pipe may look like a plastic pipe
in amplitude, whereas with an energy compensation gain (aka SEC gain), the relative reflection
strengths would be maintained.
2
Page
Figure 3 – Energy decay gain applied to data
Dewow (Bandpass Filter)
Inductive effects and antenna coupling mismatches between the radar antennas and the ground
sometimes result in very low frequency “wow” effects. These wow effects are not related to
geology or subsurface objects, and generally should be removed using a bandpass or low-cut
filter. There is no rule for the frequency range to use, since GPR systems are, by definition, ultra
wide band. A general rule-of-thumb is to use a lower cut off filter at about 0.5 f, where f, is the
published centre frequency of the radar.
Figure 4 – Data after high-pass (aka low-cut or dewow) filter
Background Removal
All radar systems, and indeed many radar environments, produce horizontal “ringing” in raw
3
datasets. As caused by the radar instrument itself, the ringing is usually due to the shielding on
Page
the antennas reverberating energy within the antenna, or internal radio frequency noise within the
electronics. There are a number of background removal types, but their purpose is to highlight
an anomaly from this constant background “noise”. This is usually done by averaging a number
of traces and subtracting that average from each trace. Care must be taken to not use too few
traces as an average or too many. Further, if there is a vertical feature of strong reflections (e.g.
the radar was moved over a metal plate), then the background removal process will create a
smearing effect near that object.
Figure 5 – Data after background subtraction
4
Page