Fishery bulletin of the Fish and Wildlife Service - NOAA

112 FISHEBY BULLETIN OF THE FISH AND WILDLIFE SERVICE
was actively in charge of the count. This method assumes that the fish are passing
over the ladders for 12 hours per day at the same average rate as observed during the
period of the count and has been applied to each ladder separately and the sum is the
estimated total for the day. For periods during which no count was made a linear
interpolation between the preceding and the following days' estimated counts was
used. Although not comparable in accuracy to the actual count, these estimates
appear to give a reasonable basis for further calculations.
A chief source of error in these counts and estimates is undoubtedly the identification
of species as the fish were passing up the ladders. After May 7, when the actual
count began, the fish were forced to pass through a small opening in a wen* placed
across each fish ladder and over a submerged platform painted white. Identification
of species under these conditions can be made with some accuracy by careful observers
and, in general, reasonable confidence can be placed in the identifications so made.
Those made under less favorable conditions must, necessarily, be accepted as the best
available. Circumstances may arise in which a particular misidentification is especially
likely to occur, in which case it may be recognized and steps taken either to
improve the identification or to determine its influence and allow for it in the estimates
of the number of fish of the species confused.
'It is apparent that one such particular case of misidentification might easily
arise during the time when the blueback run is at its peak. Gruse, which are approximately
the same size as the bluebacks, are among the chinooks and run at the same
time, and it has seemed likely that bluebacks might be mistaken for grilse or grilse for
bluebacks. An analysis has been made in which the correlation was determined between
the percentage of grilse in the total count of chinooks and the number of
bluebacks for the 10 weeks of the blueback run—June 11 to August 13.
The Pearsonian coefficient of correlation is —0.72. Using the standard procedure
the probability of chance occurrence of a coefficient of correlation as high as this is
only 0.03, so that the observed negative correlation between the percentage of grilse
and the number of bluebacks can be accepted as significant. Furthermore, it seems
likely that the relationship between these two variables is curvilinear rather than
rectilinear, as assumed by the Pearsonian coefficient, and that a true measure of the
correlation would be even higher than that calculated. Our measure is, therefore,
conservative. It seems quite likely that this negative correlation can be ascribed to
a tendency on the part of the observers to mistake grilse for bluebacks when the bluebacks
are numerous.
This raises the question as to what other errors there may be in the counts. It is
certainly difficult to distinguish species under the conditions of counting unless there
is a fairly weh 1 marked difference in size, shape, or markings, especially if light conditions
are not favorable. Observers should not be blamed for making errors under
these conditions, but, in view of the evidence of error in identification just given, it
would seem proper to investigate carefully to see how extensive these errors may be.
The importance of having properly trained and experienced observers is obvious.