14.1 Challenges to Radiological Interdiction
149
As you survey, after a fairly steady count rate of 100 cpm, you come to a
location where the count rate jumps to 110 cpm. Should you report a possible
detection?
So let’s look at the normal distribution—we can see that about 34% of the
time you should expect to see a reading between 100–110 cpm, simply due
to normal fluctuations in the count rate from background radiation. So it’s
fairly likely that this elevated reading is due to the statistical fluctuation in
normal background radiation levels and does not indicate the proximity of a
radioactive source. In fact, most of the time a “detection” is not called away
until radiation levels are at least three standard deviations above background,
and under some circumstances the threshold for a detection might be as high
as twice background levels (in this case, 200 cpm).
Now consider another example, a permanently mounted detector that
records radiation levels every second—60 data points per minute and 3600
every hour. This means that a three-sigma fluctuation in background count rate
(which happens about 0.13% of the time) will still produce 4–5 false alarms
every hour, or about 100 false alarms daily. If we increase the tolerance to
four sigma above background then about 0.05% of the counts will show up as
alarms—about two false alarms per hour.
For a single person this is a manageable level, but what about for a network
of, say, 25 detectors? If each detector is producing two false alarms each hour
then the system as a whole will be alarming 50 times an hour, or more than
1000 times daily. So setting an alarm at the five sigma level can produce an
unworkable number of false alarms with even a relatively modest number of
detectors in the field. Yet, at the same time, setting alarms at the five-sigma
level or higher can also lead to missing radioactive sources that are distant,
well-shielded, or that have weak gamma emissions.
One way to help distinguish between these possibilities is to watch radiation
levels carefully for a longer period of time to see if the slight increase drops back
to near background or if it remains consistently elevated. If a source of radiation
is actually present then the readings should fluctuate around a now-higher dose (or
count) rate; if there is no source present then radiation levels should drop again after
a short period of time and will continue to fluctuate around the expected background
radiation levels.
Another way to try to make this determination would be to travel a short distance
in one direction and observe the radiation levels, and then to repeat this in other
directions. If levels increase while traveling in one direction and drop when traveling
in the opposite direction then it is possible that a radiation source is causing these
changes. However, the radiation source could also be a granite building, a brick
building, a cemetery, or any of a number of other natural or non-threatening sources.
The person conducting the survey can also call for a large-volume scintillation
detector coupled with a multi-channel analyzer to check to see if the slightly increased
149
As you survey, after a fairly steady count rate of 100 cpm, you come to a
location where the count rate jumps to 110 cpm. Should you report a possible
detection?
So let’s look at the normal distribution—we can see that about 34% of the
time you should expect to see a reading between 100–110 cpm, simply due
to normal fluctuations in the count rate from background radiation. So it’s
fairly likely that this elevated reading is due to the statistical fluctuation in
normal background radiation levels and does not indicate the proximity of a
radioactive source. In fact, most of the time a “detection” is not called away
until radiation levels are at least three standard deviations above background,
and under some circumstances the threshold for a detection might be as high
as twice background levels (in this case, 200 cpm).
Now consider another example, a permanently mounted detector that
records radiation levels every second—60 data points per minute and 3600
every hour. This means that a three-sigma fluctuation in background count rate
(which happens about 0.13% of the time) will still produce 4–5 false alarms
every hour, or about 100 false alarms daily. If we increase the tolerance to
four sigma above background then about 0.05% of the counts will show up as
alarms—about two false alarms per hour.
For a single person this is a manageable level, but what about for a network
of, say, 25 detectors? If each detector is producing two false alarms each hour
then the system as a whole will be alarming 50 times an hour, or more than
1000 times daily. So setting an alarm at the five sigma level can produce an
unworkable number of false alarms with even a relatively modest number of
detectors in the field. Yet, at the same time, setting alarms at the five-sigma
level or higher can also lead to missing radioactive sources that are distant,
well-shielded, or that have weak gamma emissions.
One way to help distinguish between these possibilities is to watch radiation
levels carefully for a longer period of time to see if the slight increase drops back
to near background or if it remains consistently elevated. If a source of radiation
is actually present then the readings should fluctuate around a now-higher dose (or
count) rate; if there is no source present then radiation levels should drop again after
a short period of time and will continue to fluctuate around the expected background
radiation levels.
Another way to try to make this determination would be to travel a short distance
in one direction and observe the radiation levels, and then to repeat this in other
directions. If levels increase while traveling in one direction and drop when traveling
in the opposite direction then it is possible that a radiation source is causing these
changes. However, the radiation source could also be a granite building, a brick
building, a cemetery, or any of a number of other natural or non-threatening sources.
The person conducting the survey can also call for a large-volume scintillation
detector coupled with a multi-channel analyzer to check to see if the slightly increased