Downloaded 01/11/13 to 192.38.67.112. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.

org/

GEOPHYSICS, VOL. 52, NO. 11 (NOVEMBER 1987); P. 1557-1562, 6 FIGS.

Leveling airborne gamma-radiation data using

between-channel correlation information

A. A. Green*
quent flights over large lakes. Burson (1973) sampled the air
with filter systems and estimated the background from the
radiation of the dust collected. When a full gamma-ray spectrum is available, it is possible (Grasty, 1982) to estimate the
atmospheric background in the uranium channel by using the
fact that the background correlates with the intensity of the
low energy 214Pb gamma rays at 352 keY which, at normal
survey altitudes, will be due solely to the decay of atmospheric
radon.
By contrast, the potassium and thorium channels do not
suffer from background fluctuation problems to the same
extent as the uranium channel. Consequently, images of these
channels usually show little banding of the type seen in the
uranium image and, where banding does occur, it is usually
common to all channels. Artifacts of the latter type can be due
to sensitivity changes in the system, or to variable attenuation
of radiation by changing conditions of the atmosphere or the
soil moisture (Texas Instruments, 1979).
Studies of multispectral remote sensing data (Green et al.,
1987) have shown that when the signal contents of all bands
(channels) are highly correlated and when there is noise in
only one band, this noise can be removed with surprising
efficiency. Airborne radioactivity data are similar in many
ways to multispectral remote sensing images. In particular,
there is often a strong correlation among the three channels
and especially between the uranium and the thorium channels.
In the case considered here, the primary source of noise is the
background fluctuation in the uranium channel.
This paper examines the correlations among average
gamma-ray counts in the uranium, potassium, and thorium
channels for flight lines of a survey in northern New South
Wales, Australia. This information is then used to remove
flight-line dependent background errors from the uranium
channel.

ABSTRACT

A procedure for estimating background-correction

terms for the uranium channel of an airborne gammaray survey has been developed. The residuals obtained
from a multiple linear regression of flight-line means for
the uranium channel on the means for thorium and
potassium are used to correct the uranium channel for
each line. The procedure assumes that, were it not for
these background errors, the uranium flight-line means
would be a linear function of the means for potassium
and thorium. It also assumes that the background correction is the same for the whole of each line. In spite of
these limitations, the method produces good background estimates consistent with those found by more
sophisticated methods.

INTRODUCTION

The increased use of image display methods for the analysis

of airborne radioactivity data has meant that artifacts due to
incomplete or incorrect processing have become very obvious.
In particular, images of the uranium channel often show banding parallel to the flight-line direction, indicating the need for
improved background correction.
Background changes in only the uranium channel can have
a number of causes. However, the worst errors are usually
associated with the decay of atmospheric radon (Darnley and
Grasty, 1970; Grasty, 1979). The magnitude of this atmospheric background is dependent upon the weather and the
time of day. Removal of the atmospheric background has been
discussed by a number of authors. Foote (1968) used an
upward-looking detector shielded from ground radiation and
multiple reflights of some lines to estimate the atmospheric
radiation in the uranium channel. For regions with many
bodies of water, Darnley and Grasty (1970) recommended
background corrections based on counts collected during fre-

THE SURVEY DATA

Figure 1 shows the images obtained by gridding potassium

(Figure la), uranium (Figure Ib), and thorium (Figure lc)

Manuscript received by the Editor January 16, 1987; revised manuscript received May 4,1987.
.
*CSIRO Division of Mineral Physics and Mineralogy, P.O. Box 136, North Ryde, N.S.W. 2113, Austraha.
1987 Society of Exploration Geophysicists. All rights reserved.

1557

Green

Downloaded 01/11/13 to 192.38.67.112. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

1558

FIG. 1. Gridded (a) potassium, (b) uranium, and (c) thorium data for the Inverell, Grafton and Maclean 1: 250000 scale
map sheets. The images are displayed with a logarithmic contrast stretch. The uranium image shows horizontal
banding due to changing background levels.

Downloaded 01/11/13 to 192.38.67.112. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Leveling Airborne Gamma-ray Data

flight-line data. The area covered is that of the Inverell, Grafton, and Maclean 1: 250000 map sheets in New South Wales,
Australia. The detector was a 33 t crystal with a sampling
interval of I s. The nominal survey altitude was 150 m and the
flight-line spacing was 1500 m. Stripping for the Compton
scatter and corrections for aircraft altitude (Grasty, 1976) have
been applied, but background corrections could not be performed because the necessary data were not available. Data
acquired with a terrain clearance greater than 300 m were
rejected.
The data were gridded by copying the flight-line information from the located data tapes into an image file with a
625 m pixel spacing. This step resulted in an image with continuous data along the lines but with typically two or three
blank pixels between the flight lines. These blank pixels were
then interpolated using repeated passes of a 3 x 3 filter that
replaced the central pixel with the average of its nonblank
nearest neighbors. Original data remained unchanged in this
procedure. The filtering process was continued until significant
changes were no longer being made to the image. The interpolated images are displayed with a logarithmic contrast stretch
in order to enhance the low-count regions. Bright regions correspond to high count values.
The center of the study area is granite terrain which is
characterized by higher counts in all channels. In particular,
the elliptical shape of the Mole granite shows clearly. To the
east of the Mole granite, areas of low count rate correspond to
the sediments of the Clarence-Moreton basin. The regions of
very smooth image texture in all the images are those where
the interpolation procedure has had to fill large gaps in the
data because the aircraft altitude exceeded 300 m.
The uranium image (Figure lb) shows bright areas where
groups of flight lines have a significantly higher background
level than the adjacent lines. These areas have very regular
boundaries and are thus unlikely to be associated with real
geologic features. They are also enhanced when ratios are displayed. Figure 2 shows the image of the uranium-thorium
ratio computed from the data illustrated in Figure 1. The
bright areas on this image correspond to regions with flight
lines of anomalously high uranium background. Because

1559

(a)

25

a 20
QI

~ 15

10

10

20

30
40
50
Thorium line mean

60

(b)

25
c

e20
E

~ 15
E

.:2
c
e

10

:3 0
20

40

60

80

100

Potassium line mean

FIG. 3. Plots of flight-line means of the uranium channel as
functions of the (a) thorium and (b) potassium line means. The
uranium-thorium plot tends to indicate that the data are distributed with positive errors added to a line with a slope of
approximately 0.5 passing through the origin.

FIG. 2. Uranium-thorium ratio image. The banding in the uranium channel is greatly enhanced.

Downloaded 01/11/13 to 192.38.67.112. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

1560

Green

between-channel ratios are one of the most important methods of analyzing airborne radioactivity data, these errors
cause significant problems in interpretation.
Another form of leveling error also occurs in these data : a
bright horizontal band in the northeast portion of the uranium (Figure lb) and thorium (Figure lc) images. This band
does not show in the ratio image (Figure 2) and is not related
to the geology.
THE CORRECTION PROCEDURE

As noted, multichannel airborne radioactivity data are similar in many ways to multispectral remote sensing images. Both
data types depend upon the physical properties of the terrain
surface and commonly show quite strong between-channel
correlation. Thus it might be expected that processing techniques found useful for multispectral data would also be useful
in the analysis of airborne radioactivity measurements.
Recent work with airborne multispectral scanner data

(Green et aI., 1987) has led to the definition of a linear transformation that orders its new components in terms of signalto-noise ratios . In particular, for the case when noise occurs in
one band only, the noise is best isolated by the residuals of a
multiple linear regression of the noisy band on the noise-free
bands.
For the problem treated here, the essential feature of the
transformation is that it will tend to isolate background fluctuations by looking at the residuals from a regression of the
uranium channel on the other two channels. The major difference between the gamma-radiation case and the remote sensing case is that with gamma rays it is necessary to isolate noise
common to whole flight lines. The logical modification is to
perform the regression on the flight-line means for each
channel and then use the residuals to the fit as the correction
values for each flight line.
Figure 3 shows the plot of flight-line means of the uranium
channel against the means for the thorium and potassium
channels. The correlation is very high. The correlation coef-

FIG. 4. Corrected uranium channel image (with logarithmic contrast stretch). Almost all the horizontal banding has
been removed.

FIG. 5. Uranium-thorium ratio image using the corrected uranium channel data.

Downloaded 01/11/13 to 192.38.67.112. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Leveling Airborne Gamma-ray Data

ficients for the full set of flight-line means are uraniumpotassium, 0.79; uranium-thorium, 0.83; thorium-potassium,
0.90. With such high values of correlation, the multiple regression should produce a good fit with small residuals.
Here I assume that these residuals are due to background
errors associated with each flight line and that they can be
used to correct the flight-line means.
If the observed mean in the uranium channel of the ith
flight line is U, and the estimated mean is Vi' then the residual
for the ith flight line R, = U i-V i : Each flight line is then
corrected by subtracting the residual from every sample in the
line. The regression equation derived for these data is
Vi =

0.039 K i

+ 0.305 ri, + 4.4.

DISCUSSION

On first examination of the uranium image (Figure lb), it is

difficult to tell whether the lines with problems are those
brighter or darker than their surroundings. It is clear that a
problem exists, but it is unclear where the adjustments should
be made. However, because radon variations are the most

-1
Rj -2
-3

/-0.

0
-

0"\
0

-1

/\

1\\

0
0

likely cause of background variability, the errors might be

expected to be positive. This view is confirmed to some extent
by the plots of flight-line means in Figure 3, which show an
interesting distribution. The lower boundary of the uraniumthorium plot is well defined and is almost a straight line passing through the origin, but the upper boundary is diffuse. It is
as though the data were distributed on a straight line with
small, random, positive errors, as would be expected if the
radon background were fluctuating from line to line and the
average proportions of uranium and thorium were the same in
each line.
The results of applying the corrections to the uranium data
are shown in Figure 4. A significant improvement in image
quality has been achieved. The improvement is also apparent
in the uranium-thorium ratio image of the corrected data
(Figure 5). The worst artifacts have been removed from Figure
5 and the picture is much more interpretable than the original
shown in Figure 2. Note that the correction has forced the
flight lines and tie lines to match. The vertical tie-line pattern,
which is visible on the western side of the original ratio image
(Figure 2), does not appear on the corrected ratio image
(Figure 5).

1561

-2

-1

-2

~/o

V
0

3
0

j\
-2

Rj 0
-1

r\

\f\A

~1

-2

r~

IJ

-1

Cl"o-"0,
0

-2

0600

1200

0600

1200

0600

cf\

\\oJ
1200

0\
0

0
-1

-2
0600

1200

Time of day (hI

FIG. 6. Plots of residuals to the least-squares fit as a function of time of day. These are the correction terms applied to
each line. Each point represents one flight line, and each graph represents one day of flying. Residuals (RJ for only the
first few days are plotted. These residuals correspond to the lines in the southwest portion of the image where the
banding pattern is most systematic.

Downloaded 01/11/13 to 192.38.67.112. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

1562

Green

Not all the leveling problems have been corrected by this

procedure; for instance, the bright band in the northeastern
section of the image remains. This error is present in the uranium and thorium channels and thus has not been detected by
the regression. The only way to remove this type of error
without more information about the state of the system is with
a procedure that uses differences between neighboring lines to
detect lines of anomalously high counts.
The nature of the correction procedure is such that the
mean values of some flight lines will be raised and some will
be lowered, because the residuals obtained by the regression
procedure sum to zero. Although the idea of both negative
and positive corrections is somewhat at variance with the concept of positive errors due to radon, having both types of
correction is not a problem when the sole concern is the display of clean images. In effect, the data are corrected to a level
equivalent to that obtained under conditions where the radon
concentration is the average seen on this survey.
Figure 6 plots the residuals (R;) as functions of the time of
day for the first few days of the survey. On most days, the
corrections rose for the first few flight lines, and then they fell
to a lower value for the rest of the day. These curves are
similar in shape to those given by Foote (1968). His results
(obtained with an upward-looking detector) showed that the
atmospheric background first rose then feU as the day progressed. Foote ascribed this effect to the rise of the nearsurface inversion layer and its subsequent breakup in the afternoon. The fact that his correction procedure and the one
described here produce similar results gives some confidence
that this new procedure is soundly based.
The present correction procedure makes an important assumption about the relationships between the uranium, thorium, and potassium concentrations in the survey area. If
flight-line means in the uranium channel are to be linearly
related to the means in the other two channels, then on
average the relative proportions of these elements must remain
the same from line to line. The extent to which this assumption holds will be dependent upon the nature of the survey
and the variety of lithologies in each line. If different lines
occur over very different lithologies, the technique will not be
very useful. However, if normal practice in survey design is
followed, the flight lines will not be much different. Airborne
surveys are usually designed so that the flight-line direction is
across the major strike direction of the area. This means that,
more often than not, flight lines will sample a variety of lithologies in roughly similar proportions.
In situations where there are major changes in the average
proportions of elements from line to line, the correction procedure will tend to introduce leveling errors. These errors will
be obvious in the same way as the errors due to background
fluctuations. In these situations it is necessary to decide
whether the advantages obtained in improved background removal outweigh the disadvantages of the artifacts that are
introduced. In the present survey, there was a clear benefit in
the application of the correction procedure.
Another major assumption made in this work is that the
atmospheric effects are constant for the whole flight line.
Darnley and Grasty (1970) reported measurable differences in

atmospheric radioactivity at sites only a few miles apart.

Clearly, if such effects are common in the survey area, they
will limit the usefulness of this correction method. However,
when images of the data exhibit banding of the type shown in
Figure 1, corrections based on whole flight lines are likely to
improve the situation.
CONCLUSIONS

The correction procedure discussed here uses the high

between-channel correlation in airborne gamma-ray data to
remove unwanted noise that exists only in the uranium
channel. Because the potassium and thorium channels can be
expected to be very similar to the uranium channel, atmospheric interference can be detected.
The method assumes that, on the average, the proportions
of uranium, potassium, and thorium are the same on all lines
of the survey and that the atmospheric radiation is constant
for the whole of each flight line. Although these constraints
may seem severe, the correction terms this method produces
are similar to those obtained using sophisticated upwardlooking crystals, which suggests that many surveys will benefit
from using this correction procedure. The method is very
simple, involving only a trivial amount of extra computation,
and thus is worth testing when a survey has leveling problems
in the uranium channel.

ACKNOWLEDGMENTS

The support of the Australian Mineral Industries Research

Association is gratefully acknowledged. The Geological
Survey of New South Wales kindly provided the raw data and
has given permission for its publication.
REFERENCES
Burson, Z. G., 1973, Airborne surveys of terrestrial gamma radiation
in environmental research: lnst. Electr. Electron. Eng., Trans. Nucl.
Sci., NS-21, 558-571.
Darnley, A. G., and Grasty, R. L., 1970, Mapping from the air by
gamma-ray spectrometry, in Proc., 3rd Internat. Geochemical
Sympos., Toronto: Can. Inst. Min. Metall. Special Vol. 11,485-500.
Foote, R. S., 1968, Improvement in airborne gamma-radiation data
analysis by removal of environmental and pedologic radiation
changes, in Sympos. on the Use of Nuclear Techniques in Prospecting and Development of Mineral Resources: Internat. Atomic
Energy Agency Mtg., Buenos Aires, 187-196.
Grasty, R. L., 1976, A calibration procedure for an airborne gammaray spectrotneter: Geol. Surv. Canada Paper 76-16.
- - - 1979, Gamma-ray spectrometric methods in uranium
exploration-Theory and operational procedures, in Hood, P. J.,
Ed., Geophysics and geochemistry in the search for metallic ores:
Geol. Surv. Canada, Economic Geol. Rep. 31,147-161.
- - - 1982, Utilizing experimentally derived multichannel gammaray spectra for the analysis of airborne data, in Proc., Sympos. on
Uranium Exploration Methods, Review of the Nuclear Energy
Agency R & Dzlntemat. Atomic Energy Agency Programme, Paris,
653-669.
Green, A. A., Berman, M., Switzer, P., and Craig, M. D., 1987, A
transformation for ordering multispectral data in terms of image
quality with implications for noise removal: Inst. Electr. Electron.
Eng., Trans. Geosci. Remote Sensing, in press.
Texas Instruments Inc., 1979, Study of airborne gamma-ray spectrometer data procedures Wind River basin, Wyoming: Final rep.
for U.S. Dept. of Energy GJBX-40.