F. Brakstad et al., PREDICTION OF THE SHANNON WIENER DIVERSITY INDEX FROM TRACE-ELEMENT PROFILES IN SEDIMENTS AROUND THE STATFJORD PLATFORMS, Chemosphere, 29(7), 1994, pp. 1441-1465
A strategy for building multivariate models capable of predicting chan
ges in community structure from environmental data is presented This i
ncludes five important elements in the process of model building and v
alidation concerned with pollution surveys; i) quality control of chem
ical replicates, ii) data exploration with the purpose of detecting di
fferent distribution profiles of pollutants, iii) transformation to ac
hieve constant variance of the environmental data in order to improve
the models' predictive ability, iv) validation of the models in time a
nd space, and, v) multivariate testing for outliers among new samples
prior to prediction by the model. This methodology is applied to the s
ediments around the three Statfjord fields in the North Sea, Three ind
ependent models based on the trace element concentrations of selected
training samples, explain in average 86% of the variation of the Shann
on Wiener diversity index. Accurate predictions of the diversity index
were achieved on test samples beyond distances included in the traini
ng set, and on samples collected two years later. High concentrations
of selected trace elements, in particular Cu, correlate strongly with
low diversity.