Geology Reference
In-Depth Information
1.5.5 Repeatability
Repeat data are vital for checking whether an instrument is performing to
specification. Ideally, a repeat survey line should be completed on every
survey grid before moving to the next grid. For linear transects or meander-
ing surveys, a minimum of 5% of repeat data is required. Repeat line-data
achieve two things - they confirm that the instrument is responding consis-
tently and they also provide a measure of the positioning accuracy. Where
geophysical anomalies are small, it may be prudent to collect more than one
repeat line per survey grid, because of low signal-to-noise ratios. In grav-
ity surveys requiring microGal resolution, it may be necessary to reoccupy
two or more stations in each loop. Repeatability requirements should be
discussed and agreed with the client before a survey begins.
1.5.6 Detection limits
To a geophysicist, signal is the object of the survey and noise is anything else
that is measured but is considered to contain no useful information. Using
geophysics to locate a target is in some ways analogous to receiving a mobile
phone message. If the ratio of signal to noise is high (good 'reception'), a
target may be found at close to the theoretical limits of detection. If the signal
is weak it may not be possible to distinguish enough of the 'conversation' to
make it understandable, or the 'connection' may be lost completely. 'Made'
ground often contains material that interferes with the geophysical signal,
so that the signal-to-noise ratio may be low even though the signal is strong.
It may then not be possible to distinguish the target.
One observer's signal may be another's noise. The magnetic effect of
a buried pipe is a nuisance when interpreting magnetic data in geological
terms but may be invaluable to a site developer. Much geophysical field
practice is dictated by the need to improve signal-to-noise ratios. In many
cases, as in magnetic surveys, variations in a background field are a source
of noise and must be precisely monitored.
1.5.7 Variance and standard deviation
The statistics of random noise are important in seismic, ground radar, radio-
metric and induced polarisation (IP) surveys. Adding together N statistically-
long random series, each of average amplitude A , produces a random series
with amplitude A × N .Since N identical signals of average amplitude A
treated in this way produce a signal of amplitude A × N , adding together
( stacking ) N signals containing some random noise should improve signal-
to-noise ratios by a factor of N .
Random variations may have a normal or Gaussian distribution, produc-
ing a bell-shaped probability curve. A normal distribution can be charac-
terised by a mean (equal to the sum of all the values divided by the total
Search WWH ::




Custom Search