What shall we teach in environmental statistics?

The need for statistical methodology in environmental and ecological applications has grown dramatically over the past few decades, to where targeted and/or specialized courses in environmental statistics are necessary at both the undergraduate and graduate levels. We discuss here the construction of such courses, and pose questions on the course development process for the statistical and environmental community. Our exposition is based upon our own experience with the design of a graduate environmental statistics course.     

Environmetrical interpretation of analytical data of marine organisms from the black sea

The environmetrical analysis of the benthic organisms data set from a Black sea region has revealed new information concerning the chemical content and the bioindicating abilities of polychaeta (Melina palmata), Crustacea (Aspendopsis ostroumovi) and molluscs (Mytilus gallo‐provincialis). The application of various multivariate statistical approaches like cluster and principal component analysis, linear regression and partial least square modeling, source apportioning makes it possible to understand in a better way the properties of the benthic organism as collectors of pollutants in a total and a more specific mode. It is shown that heavily polluted coastal zones are indicated in the same way by all benthic species but some specificity could be detected when moderately polluted zones are considered. In this case polychaeta accumulated preferably Co, Cr, Cu and Pb Crustacea ‐ As, Cd and Ni and molluscs ‐ Zn to a limited extent.

PCA identified three latent factors (“anthropogenic flotation”;, “anthropogenic galvanic”; and “naturally occurring") which explain about 65% of the total variance of the system and determine the data set structure. The source apportioning on the absolute principal component scores proved that none of the metals is quantitatively linked with only one anthropogenic or natural source.

The linear regression and PLS models have indicated that a reliable prognosis of the pollution on some naturally occurring chemical components (e.g. linear regression on Zn for Cr and Ni) or combination of them (PLS modeling on Mn/Zn or on Mn/Zn/Fe for the rest of the pollutants) could be achieved.