RAINOUT RELATIONS ON SAMPLING NETWORKS
517
were positive on the larger 6000 square mile network. With duration
and rate considerable fluctuation from negative to positive coefficients
occurred with both concentration and deposition.
All available data from the rainwater sampling network were used
in the plots shown in Fig. 3. However, only the East Central Illinois
network of 400 square miles was inoperation during the August storms.
The eight storms described in Fig. 3 were selected to illustrate typical
correlations between radioactive rainout and rainfall volume at a
group of points in a relatively small area in the same convective
storm. In some storms, such as those ofJune10, July 1, and August 28,
there was a pronounced trend for the beta concentration to vary in-
versely with the rainfall amount. Correlation coefficients of —0.82,
—0.76, and —0.77, respectively, were obtained in these three storms.
In other storms described in Fig. 3, however, the inverse trend was
relatively weak, and, in the storm of August 6, the trend was reversed
to produce a correlation coefficient of 0.58. Several investigators,
employing storm-to-storm data at a single point, have noted an inverse correlation between radioactivity per unit volume and rainfall
amount. The results of the Illinois study, in which multiple sampling
points within the same storm were employed, lead to the conclusion
that an inverse correlation usually prevails from point to point in
convective storms, but the correlation is often weak. Occasionally, it
is positive rather than negative.
An effort was made to determine whether the correlation between
radioactivity and rainfall would improve if the storm rainfall were
divided into several parts instead of using totalstorm data. Convective
storms usually consist of several rain bursts, and it is possible that
combining all the storm rainfall into a single sample masks the true
radioactivity—rainfall relation to some extent. Automatic rainwater
Samplers used at 16 stations in 1963 provided consecutive samp!
containing 0.04 to 0.06 in., and up to 12 samples were obtained in eacn
storm at stations in the automatic-sampler network.?
For each of 20 storms in which a large number of these samples
were obtained within the network, graphical plots were made in which
gross beta concentration was related to rainfall rate. The scatter of
points in most storms wasrelatively great and indicated a low degree
of correlation. For example, in the storm of June 10, a total of
60 samples was obtained from 12 stations, which showed a correlation
coefficient of —0.35 between beta concentration and rainfall rate. Since
the graphical plots indicated that this storm provided one of the better
relations among the 20 storms investigated, no further mathematical
calculations were made. In the 20 storms the trend found mostfrequently was for the beta concentration to decrease with rainfall rate,
but the opposite tendency occurred in several storms.