Download (20MB) - Repository@Napier

There are many sample designs which can be employed to assess the spatial patterns
of marine benthic populations and communities, each of them being more or less
suitable for different habitats and different scales (Andrew and Mapstone, 1987).
Contiguous sample designs give less equivocal assessments of spatial patterns since
the abundances of all individuals within the extent of the survey are determined.
When the extent of the survey is too large for contiguous sample designs, assessment
of spatial patterns becomes more inexact.
McArdle and Blackwell (1989) proposed that a systematic (grid) sample design was
the most suitable for the detection of spatial patterns. This design will usually provide
the smallest estimate of the mean (Milne, 1959) and patterns can be detected and
displayed easily. The present results suggest that when the lag is large relative to the
average patch size, however, this design may lead to equivocal results. Grid sample
designs were used in this study to determine the spatial patterns of P. elegans on
Drum Sands. Using mapping and spatial autocorrelation analysis, the size and spatial
distribution of P. elegans patches could only be determined using a survey in which
the lag, or distance between samples, was less than the mean size of the patches and
less than the average distance between patches. Consequently, the 1 m survey
provided important information about the size and spatial arrangement of small-scale
P. elegans patches. The 8m and 40m surveys, however, revealed little information
about the spatial distribution of P. elegans because the results obtained from these two
surveys depended upon the actual spatial arrangement of the small-scale P. elegans
patches. For example, if two neighbouring plots were located on separate P. elegans
patches in either the 8m or the 40m surveys, mapping and spatial autocorrelation
analysis may have erroneously indicated the presence of larger scale patches. This
was revealed by the present study because P. elegans is a tube-building polychaete
and, therefore, high-density areas were very visible on the sandflat. For subsurface
invertebrate species, or species in which high-density areas are less readily observed, a
grid-sample design can potentially lead to misleading results when patches smaller in
size than the lag are present. For such species, a hierarchical sampling design may be
more suitable since patchiness at a larger range of scales can be detected
simultaneously (Morrisey et al., 1992). This sample design lacks the resolving power
of grid-sample designs and, therefore, does not give information about the exact size
236