Geophysical data acquisition, processing and interpretation - Geophysics for the Mineral Exploration Geoscientist - page 33

data for enhancement and display. It is essential that, as far
as is possible, all acquisition-related errors, and errors and
noise emanating from external sources, be corrected or
compensated as a first stage of data processing. Otherwise,
the errors will propagate and increase through the various
stages of data processing. The situation is stated succinctly
by the maxim:

2.7.1.2 Levelling
The success of the data reduction in eliminating errors can
be assessed by comparing corrected repeat measurements
made at the same location. Ideally the reduced (repeat)
values should be identical. Any remaining differences,
known as residual errors, can be used to further reduce
the data to remove these errors. This process is not based
on measurements from the natural environment; instead it
is a pragmatic attempt, based on statistical methods, to
lessen the in uence of remaining errors by redistributing
them across the entire dataset. The process is known as
levelling because the adjustment made to the amplitude of
each measurement changes the overall amplitude level of
the measurements.
Tie lines provide the necessary repeat readings where
they intersect the survey lines (see Section 2.6.3.3 ). The
basic idea is to use the tie-line data to

.
Reduction compensates for various sources of error and
noise, sometimes using secondary data acquired during the
survey (see Section 2.6 ). Some reductions may be carried
out automatically by the data acquisition instrumentation,
provided that the instrument records the necessary second-
ary data and that the magnitude of the correction is easily
determined. More complicated reduction processes are
applied post-survey.
In its simplest form, reduction involves a manual assess-
ment of the data to remove readings that are obviously
unsuitable, for example those dominated by noise. Correc-
tions applied during reduction include compensation for
the geophysical response of the survey platform; variable
sensor orientation/alignment; and various temperature,
pressure and instrumental effects. In addition, corrections
are applied to suppress environmental noise (see Table 2.1 ) ,
which may require measurements obtained from a second-
ary sensor speci cally deployed to monitor the noise. The
corrections applied during reduction of the data are only as
good as the secondary data on which they are based. For
example, information about the local topography may be
insufficient to remove its effects completely.
Generally, reduction processes are parameter specific
and survey type specific, and are described in detail later
in our descriptions of each geophysical method. Here we
describe some more generally applicable operations.

' Errors don ' t go away, they just get bigger '

adjacent survey
lines together at regular intervals, with the residual errors
at the line intersections used to adjust the data. For air-
borne surveys this is an imperfect process because of the
inevitable errors due to differences in survey heights at the
line intersections, i.e. the two readings are not from exactly
the same location. Tie lines may also be part of a ground
survey; but if measurements have been made at one or
more base stations, then these provide the necessary repeat
readings. It is common practice to make repeat readings at
selected stations for this speci c reason.
There are various means of redistributing the residual
errors; see Luyendyk ( 1997 ), Mauring et al.( 2002 ) and Saul
and Pearson ( 1998 ) for detailed descriptions regarding
airborne datasets. One simple approach when tie lines are
available is based on using the residual values at the inter-
sections to interpolate a 2D error function across the
survey area. By subtracting the relevant error function
values from the readings along the survey lines,

' tie '

the

2.7.1.1 De-spiking
Spikes, or impulses, are abrupt changes in the data which
have short spatial or temporal extent, and usually occur as
a single data point. These may be caused by instrumental
problems or they may originate in the natural environ-
ment. Spikes can be removed using various numerical data
processing techniques (see Smoothing in Section 2.7.4.4 ) or
by manual editing. Erroneous data are then replaced by
interpolating from adjacent readings. It is important to
de-spike early in the processing sequence since some pro-
cessing operations will produce artefacts if spikes are not
removed, notably those that involve a Fourier transform
(see Appendix 2 ).

residual errors are reduced.
A quick and easy way to assess the quality of the data
reduction is to view it with shaded relief (see Section
2.8.2.3 ) , with the illumination perpendicular to the survey
line direction, optimum for highlighting across-line level-
ling errors. Residual errors in the levelled data are
revealed as corrugations or ripples between the survey
lines (see Fig. 3.24 ) . Also, derivative images, being sensi-
tive to gradients, are usually effective in revealing residual
errors in a levelled dataset. Any remaining errors indicate
either that the various corrections applied are imperfect or
that other

sources of survey error have not been

accounted for.