SOIL ARTHROPOD SAMPLINQ
7
Work of this kind has not yet been done for soil animals but it should
be noticed that since only presence is being recorded very small sampleunits should be possible - and indeed required, so as to make certain
that a proportion of units will be without animals. Under some circumstances this might facilitate taking the large numbers of sample-units
which such an analysis demands.
A different approach to community analysis has been suggested -
and illustrated - by Fager (1957) who records for each sample-unit the
rank order (i.e. the order of dominance) of each species. With the aid of
rank correlation analysis developed by Kendall (1955) the affinity of
different sample groups can be determined and standard errors can be
estimated. Again, the size of sample-units can be considerably reduced
compared with those required for complete counts and statistical significance of the results can be established. Also the counting labour can
presumably be lessened because a relatively cursory inspection will often
suffice to establish order of dominance. When samples regularly contain
large numbers of small animals and smaller numbers of large ones, for
example oribatid and parasitid mites, it will be desirable to analyse each
main group separately.
Sequential analysis [for practical details of calculations see Waters
(1955)l is another technique not yet applied to community studies
of soil, but which appears to offer substantial savings in labour, and
which demands relatively small samples. In cases where a particular
species is already established as a valuable indicator and sets of sampleunits are being examined for affinity in terms of this species, sequential
methods should provide a means of reducing the number of sampleunits examined to a minimum, although it is unlikely to reduce substantially the number of units taken from the soil in the first place. A
somewhat similar method was developed by Capstick (1959) for detecting significant differences between aliquot samples with minimum
labour.
A final point to be remembered in connection with the size of sampleunits in community studies is that when they are made too small the
rarer species are likely to be excluded altogether. It is frequently observed that the commoner species are not those whose distribution is the
most useful indicator of faunistic groupings and, as shown by Hairston
(1959) it is important to include the less common species. It seems,
therefore, that any attempt a t a complete community analysis may
require either parallel samples with units of different sizes or else the
collection of rather numerous sample-units for the rare species and the
use of aliquot parts of these or a sequential analysis scheme for obtaining statistics on the commoner ones.
The importance of deciding a t the outset whether or not absolute
7
Work of this kind has not yet been done for soil animals but it should
be noticed that since only presence is being recorded very small sampleunits should be possible - and indeed required, so as to make certain
that a proportion of units will be without animals. Under some circumstances this might facilitate taking the large numbers of sample-units
which such an analysis demands.
A different approach to community analysis has been suggested -
and illustrated - by Fager (1957) who records for each sample-unit the
rank order (i.e. the order of dominance) of each species. With the aid of
rank correlation analysis developed by Kendall (1955) the affinity of
different sample groups can be determined and standard errors can be
estimated. Again, the size of sample-units can be considerably reduced
compared with those required for complete counts and statistical significance of the results can be established. Also the counting labour can
presumably be lessened because a relatively cursory inspection will often
suffice to establish order of dominance. When samples regularly contain
large numbers of small animals and smaller numbers of large ones, for
example oribatid and parasitid mites, it will be desirable to analyse each
main group separately.
Sequential analysis [for practical details of calculations see Waters
(1955)l is another technique not yet applied to community studies
of soil, but which appears to offer substantial savings in labour, and
which demands relatively small samples. In cases where a particular
species is already established as a valuable indicator and sets of sampleunits are being examined for affinity in terms of this species, sequential
methods should provide a means of reducing the number of sampleunits examined to a minimum, although it is unlikely to reduce substantially the number of units taken from the soil in the first place. A
somewhat similar method was developed by Capstick (1959) for detecting significant differences between aliquot samples with minimum
labour.
A final point to be remembered in connection with the size of sampleunits in community studies is that when they are made too small the
rarer species are likely to be excluded altogether. It is frequently observed that the commoner species are not those whose distribution is the
most useful indicator of faunistic groupings and, as shown by Hairston
(1959) it is important to include the less common species. It seems,
therefore, that any attempt a t a complete community analysis may
require either parallel samples with units of different sizes or else the
collection of rather numerous sample-units for the rare species and the
use of aliquot parts of these or a sequential analysis scheme for obtaining statistics on the commoner ones.
The importance of deciding a t the outset whether or not absolute