METHODS OF SAMPLING THE BENTHOS
247
account of their abundance, while Longhurst’s recruitment curve
relates to the densities of the species taken.
The straight-line relationship described by Williams seems to hold
reasonably well for the different benthic populations to which it has
been applied. It no doubt depends on the existence of a continuous
spectrum of species of differing population densities, and might prove
unreliable under certain conditions where this is not so.
Sampling at sea is random in that the exact spot on which the
instrument is lowered cannot be predetermined, but the spacing of
successive samples will depend on the handling of the ship. If the
ship is anchored a whole series may be taken as the ship swings to the
anchor. The possibility of sampling twice in exactly the same place
will only occur when working from a fixed point such as a raft or an
ice hole. More often the ship is drifting, and on some grounds each
successive haul will be so different that any st,atistical comparison of
the samples would be unwise. It may then only be possible to take
one sample per station.
Series of samples on a reasonably homogeneous ground may be
treated by the usual statistical techniques. The closeness to a Poissontype distribution may be assessed by the Coefficient of Dispersion, as
used for example by Salt and Hollick (1946) for wire-worm populations.
The coefficient :
Z ( z - 3 ) a
z ( n - 1)
is based on the fact that thc Variance is equal to the Mean in the
Poisson distribution. This coefficient is equal to unity in a randomly
dispersed population, to’ more than one if there is a tendency to
aggregation, and to less than one if the distribution is more even than
random. The terms over- and under-dispersed as applied to evenly
distributed and aggregated populations respectively are confusing and
should not be used. The test of significance for deviation from
random is :
____ -
-
(n -
where n is the number of samples.
The coefficient of dispersion formula has been used for benthos
studies by Holme (1950, i953); Clark and Milne (1955), Ohba (1969),
Ursin (1960) and M. L. Jones (1961). However, Ursin (1960) has shown
that in my own results (Holme, 1953) the coefficient of dispersion for
the commoner species in the sample is greater than for the less common
ones. He suggests that this may be partly due to “technical
h .I .B.-2
247
account of their abundance, while Longhurst’s recruitment curve
relates to the densities of the species taken.
The straight-line relationship described by Williams seems to hold
reasonably well for the different benthic populations to which it has
been applied. It no doubt depends on the existence of a continuous
spectrum of species of differing population densities, and might prove
unreliable under certain conditions where this is not so.
Sampling at sea is random in that the exact spot on which the
instrument is lowered cannot be predetermined, but the spacing of
successive samples will depend on the handling of the ship. If the
ship is anchored a whole series may be taken as the ship swings to the
anchor. The possibility of sampling twice in exactly the same place
will only occur when working from a fixed point such as a raft or an
ice hole. More often the ship is drifting, and on some grounds each
successive haul will be so different that any st,atistical comparison of
the samples would be unwise. It may then only be possible to take
one sample per station.
Series of samples on a reasonably homogeneous ground may be
treated by the usual statistical techniques. The closeness to a Poissontype distribution may be assessed by the Coefficient of Dispersion, as
used for example by Salt and Hollick (1946) for wire-worm populations.
The coefficient :
Z ( z - 3 ) a
z ( n - 1)
is based on the fact that thc Variance is equal to the Mean in the
Poisson distribution. This coefficient is equal to unity in a randomly
dispersed population, to’ more than one if there is a tendency to
aggregation, and to less than one if the distribution is more even than
random. The terms over- and under-dispersed as applied to evenly
distributed and aggregated populations respectively are confusing and
should not be used. The test of significance for deviation from
random is :
____ -
-
(n -
where n is the number of samples.
The coefficient of dispersion formula has been used for benthos
studies by Holme (1950, i953); Clark and Milne (1955), Ohba (1969),
Ursin (1960) and M. L. Jones (1961). However, Ursin (1960) has shown
that in my own results (Holme, 1953) the coefficient of dispersion for
the commoner species in the sample is greater than for the less common
ones. He suggests that this may be partly due to “technical
h .I .B.-2