Home
The French Polynesian Atolls Fundamentals of Reef Ecology The Tuamotu atoll Communities
Home     Glossary  
  Fluxes of Matter in the Tuamotu atolls Types of atoll and the Ecosystems Natural Resources and their Management
Effects on the water column Effects on the benthos
  and the fish 
Effects on the trophic food web and the natural resources

 

Methods used to analyse the relationships between the morphological and functional characteristics of an atoll

R. Claereboudt, P. Dufour

 

Multi Dimensional Scaling (MDS)

 

With the Multidimensional Scaling Method developed by Shepard (1962) and Kruskal (1964), it is possible to  draw up a spatial representation (or map) of a set of samples (in this case, lagoons) by expressing the distances between the samples in terms of a similarity matrix based on all the internal and forced variables involved (IVs and FVs,  respectively). In other words, if sample a is more similar to sample b than to sample c, a will be placed nearer to b than to c on the spatial map.

 

 Links between forced variables and internal variables  
based on Clarke and Ainsworth’s method.

 

With the method developed by Clarke and Ainsworth (1993), the similarity matrixes between the samples (in this case, the lagoons) are accounted for by the internal descriptors (the IVs) and the environmental descriptors (the FVs). This method is based on a fairly logical principle, according to which if the series of FVs on which the structure of the system depends is known, a similar pattern of resemblance between the atolls should emerge to that calculated using the IVs. In other words, the map based on the IVs should be similar to that based on the FVs. If any of the main FVs. are omitted, the two maps will match less satisfactorily. Likewise, upon introducing the FVs, which do not contribute to the internal structure of the lagoons, the degree of similarity between the two maps will again deteriorate and there will be some background noise.

In the framework of the present analysis, the fit between the two similarity matrices based on the FVs and the IVs was expressed in terms of a matrix correlation coefficient. This coefficient is equal to zero if the two matrices compared are completely dissimilar and equal to 1 if they are identical. A matrix correlation coefficient greater than 0.5 definitely indicates the existence of some link between the FV and IV maps, whereas the significance of a coefficient of only 0.4 or 0.3 is difficult to assess.

 

 

 

update : 07/10/08

Search

Atoll_site_webmaster