Counting butterflies
Michael Healy
MJRHEALY at compuserve.com
Thu Jul 31 13:13:45 EDT 1997
-------------- next part --------------
All transect-walkers - or perhaps only number-obsessionals like me - ask
themselves how repeatable their counts are, what would happen if, at the end
of a walk, they went round again. To throw some light on this, I recently
walked a 2 km stretch of disused railway line (the Nicky Line at Harpenden,
Hertfordshire, between Tl124147 and TL114135) five times in each direction
on a clear sunny day between 1030 and 1445. The results were as follows -
Trip no 1 2 3 4 5 6 7 8 9 10
Start time 1027 1051 1117 1142 1208 1233 1300 1325 1349 1412
Temperature (C)24 26 25 26 25 25 25 24 23 24
T sylvestris
/lineola 0 0 1 1 2 2 1 2 0 5
O venata 0 2 8 3 6 2 1 4 6 4
P brassicae 3 4 4 3 6 9 3 6 2 5
P rapae 24 24 19 27 27 28 22 31 30 21
P napi 5 3 2 2 6 6 4 5 1 7
A urticae 1 2 4 2 0 2 2 0 1 1
P tithonus 1 7 5 9 4 10 7 8 6 5
M jurtina 6 11 14 10 13 12 17 8 10 12
Total 40 53 58* 57 64 71 58* 65* 57* 60
*Plus singletons -
Trip 3 P icarus
Trip 7 P aegeria
Trip 8 V atalanta
Trip 9 M galathea
The first trip gives a somewhat lower count than the others, in accordance
with the standard instruction not to undertake transect walks before 1100.
There is a slight tendency for the even-numbered northbound trips to give
higher counts than the southbound ones - the line runs approximately
north-south and it is easier to count the butterflies with the sun at your
back. Apart from this, the counts, both for the total and for the commoner
species, are reasonably consistent.
It is in fact, using a rather simplistic statistical argument, possible to
make a rough estimate of the fraction of the total population which is
included in the counts, assuming this to be constant over the period. If
this fraction is very small, statistical theory suggests that the variance
of the counts (the square of the standard deviation) is expected to equal
the mean. If all the butterflies are seen on every occasion, all the counts
will be the same and the variance will be zero. Between these extremes the
expected variance is given by the mean times (1 - f) where f is the fraction
observed. Applying this argument to the overall totals and to the counts
for P rapae and M jurtina (omitting trip no 1) gives f = 0.51, 0.32 and 0.42
respectively, suggesting that the total population might be between two and
three time the numbers actually observed. Even apart from the over-simple
statistical model, these figures should not be taken too seriously, since an
estimated variance based on no more than 9 observations is necessarily
very imprecise.
It is of some interest that a regular transect walk the following day, under
similar weather conditions, gave a total count of 87, including 37 P rapae
and 20 M jurtina. On the evidence, this represents a real increase in the
population over 24 hours.
Michael Healy
31 July 1977
More information about the Leps-l
mailing list