18.08.2013 Views

Download (20MB) - Repository@Napier

Download (20MB) - Repository@Napier

Download (20MB) - Repository@Napier

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

neighbouring samples usually assume similar values, may arise from gradients or<br />

patches exceeding the diameter of inter-sample distances. Significant positive<br />

autocorrelation (high Moran's i and low Geary's c) at distance class 1 therefore<br />

implies a patch size of between 1-1.5m2 and at distance class 2, a patch size of 1.5-<br />

2m2.<br />

Generally, there were very good agreements between the two autocorrelation<br />

coefficients used: Geary's coefficient was nearly always significant when Moran's<br />

coefficient was significant and both gave similar estimates of patch sizes in most<br />

cases. Cliff and Ord (1973) proposed that i was generally better than c, although the<br />

margin of the advantage was slight. Therefore, patch sizes are given here as indicated<br />

by Moran's coefficient.<br />

The density plots (Figures 2.5(i-vii)) suggest that all of the species with significantly<br />

non-random distributions formed areas of increased densities, or patches, of either less<br />

than 1m2 (e.g., M. balthica and E. cf flava) or between 1-2m 2 (e.g., P. elegans, A.<br />

marina, L. conchilega, G. duebeni and B. sarsi) with very low abundances outside the<br />

patches. However, mapping only gives a guide to spatial patterns, they do not confirm<br />

them. The patterns indicated in Figures 2.5(i-vii) are supported by their correlograms.<br />

For example, patch sizes smaller than inter-sample distances (or lag, in this case 1m)<br />

are not revealed by spatial autocorrelation analysis and therefore M. balthica and E. cf<br />

flava had non-significant correlograms. For these species, smaller inter-sample<br />

distances would have been needed to determine their underlying spatial structure. The<br />

significant correlograms with positive autocorrelation at distance class 1 indicate that<br />

P. elegans, A. marina, L. conchilega and G. duebeni formed patches of 1-1.5m2 while<br />

the positive autocorrelation at distance class 2 suggests that B. sarsi formed patches of<br />

1.5-2m2. For P. elegans, 2 patches were sampled wholly within the 8x8m grid, with 3<br />

others on the edge. These corresponded to the positions of areas of smooth, raised<br />

sediment (pers. obs.) and therefore confirm that these areas did indeed contain<br />

significantly increased numbers of P. elegans. Furthermore, Figure 2.5(i) shows that<br />

although the patch sizes were approximately equal, the inter-patch distances varied<br />

suggesting that the P. elegans patches were randomly distributed at this scale.<br />

37

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!