S 0 I I, A RTH R 0 1’0 1) S A A1 1’LI N G
n
Taylor (1961) claims that biological data are better fitted to a power
series than to the negative binomial. This is given by s2 = anab (where
s2 =variance, m =mean, a = constant which varies with sample size,
b=constant which varies with, and can be used to measure, aggregation). He suggests tests of significance and claims that b is a biologically
valid measure of aggregation.
There remains the physical problem of how the sample-units should
be disposed on the ground; whether a t random, on a regular lattice, on
transect lines or a combination of such patterns. Once again the purpose
of the study is the determining factor. A trophic study based on a single
experimental area can be gridded and sample-units taken a t random. A
rather more efficient procedure which still avoids bias is to use stratified
random sampling; the whole area being equally divided into sub-areas
or “strata” and equal numbers of sample-units being taken at random
from each stratum. A linear arrangement of strata can also be used in a
community type study when distribution along a catena or transect line
is to be investigated, but usually, it is more practicable to select a number of areas along the transect which are gridded and sample-units are
then taken randomly from each. Under the crudest field conditions
when soil faunas are to be related to floristic or other discernible
regions there may be insufficient time to construct grids and to sample
at random and in this case the best compromise seems to be to sample
at the places where an object thrown into the air lands. In this way unconscious bias is largely avoided.
The completely regular grid has the advantage over random sampling
in that efficiency is higher but there is the remote possibility that
sample-points will resonate with natural features. Of course when a
complete map of spatial distribution is required a rigid grid with
sample-units from each square should be used.
To conclude this brief survey of statistical aspects of soil arthropod
sampling, it seems appropriate to stress that such methods allow the
ecologist who takes the trouble to formulate clear-cut hypotheses to plan
an efficient programme of work. There is a wide range of methods appropriate to different types of study and by careful selection the ecologist
can not only obtain significant answers with a minimum of labour but
also, because he is forced to concentrate on principles he can finally satisfy
the natural historian’s just complaint that so much quantitative ecology
has been mechanically applied and divorced from biological reality.
B. LABORATORY METHODS FOR COVNTING AND RECORDING
Quite apart from questions of sampling in the field and the actual
means by which the soil fauna is removed from the sample-units, a good
deal of variation exists in the methods by which the animals are counted
A 2
1 . R . - I
n
Taylor (1961) claims that biological data are better fitted to a power
series than to the negative binomial. This is given by s2 = anab (where
s2 =variance, m =mean, a = constant which varies with sample size,
b=constant which varies with, and can be used to measure, aggregation). He suggests tests of significance and claims that b is a biologically
valid measure of aggregation.
There remains the physical problem of how the sample-units should
be disposed on the ground; whether a t random, on a regular lattice, on
transect lines or a combination of such patterns. Once again the purpose
of the study is the determining factor. A trophic study based on a single
experimental area can be gridded and sample-units taken a t random. A
rather more efficient procedure which still avoids bias is to use stratified
random sampling; the whole area being equally divided into sub-areas
or “strata” and equal numbers of sample-units being taken at random
from each stratum. A linear arrangement of strata can also be used in a
community type study when distribution along a catena or transect line
is to be investigated, but usually, it is more practicable to select a number of areas along the transect which are gridded and sample-units are
then taken randomly from each. Under the crudest field conditions
when soil faunas are to be related to floristic or other discernible
regions there may be insufficient time to construct grids and to sample
at random and in this case the best compromise seems to be to sample
at the places where an object thrown into the air lands. In this way unconscious bias is largely avoided.
The completely regular grid has the advantage over random sampling
in that efficiency is higher but there is the remote possibility that
sample-points will resonate with natural features. Of course when a
complete map of spatial distribution is required a rigid grid with
sample-units from each square should be used.
To conclude this brief survey of statistical aspects of soil arthropod
sampling, it seems appropriate to stress that such methods allow the
ecologist who takes the trouble to formulate clear-cut hypotheses to plan
an efficient programme of work. There is a wide range of methods appropriate to different types of study and by careful selection the ecologist
can not only obtain significant answers with a minimum of labour but
also, because he is forced to concentrate on principles he can finally satisfy
the natural historian’s just complaint that so much quantitative ecology
has been mechanically applied and divorced from biological reality.
B. LABORATORY METHODS FOR COVNTING AND RECORDING
Quite apart from questions of sampling in the field and the actual
means by which the soil fauna is removed from the sample-units, a good
deal of variation exists in the methods by which the animals are counted
A 2
1 . R . - I