Michele Scardi

Applications of self-organizing maps for ecomorphological investigations through early ontogeny of fish.

We propose a new graphical approach to the analysis of multi-temporal morphological and ecological data concerning the life history of fish, which can typically serves models in ecomorphological investigations because they often undergo significant ontogenetic changes. These changes can be very complex and difficult to describe, so that visualization, abstraction and interpretation of the underlying relationships are often impeded. Therefore, classic ecomorphological analyses of covariation between morphology and ecology, performed by means of multivariate techniques, may result in non-exhaustive models. The Self Organizing map (SOM) is a new, effective approach for pursuing this aim. In this paper, lateral outlines of larval stages of gilthead sea bream (Sparus aurata) and dusky grouper (Epinephelus marginatus) were recorded and broken down using by means of Elliptic Fourier Analysis (EFA). Gut contents of the same specimens were also collected and analyzed. Then, shape and trophic habits data were examined by SOM, which allows both a powerful visualization of shape changes and an easy comparison with trophic habit data, via their superimposition onto the trained SOM. Thus, the SOM provides a direct visual approach for matching morphological and ecological changes during fish ontogenesis. This method could be used as a tool to extract and investigate relationships between shape and other sinecological or environmental variables, which cannot be taken into account simultaneously using conventional statistical methods.

“Right” or “wrong”? insights into the ecology of sidedness in european flounder, Platichthys flesus

Sidedness polymorphism in flatfish has been linked to ecological selection between morphs. However, the alternate hypothesis that morphological differences between right‐ and left‐sided forms may be due to errors during development, as a consequence of disturbed homeostasis, which still remains largely unexplored. Here, we examined the case of Platichthys flesus (flounder), a polymorphic flatfish exhibiting large and clinal variation in the frequency of the left‐sided morph, which is the reversed condition in this generally right‐sided species. An integrated approach consisting of the analyses of shape variation, stomach contents, and skeletal anomalies was used. Morphological differences were observed between morphs, which are in agreement with previous findings in a congeneric species (Platichthys stellatus). In parallel, significant differences in feeding choices were detected, suggesting a coherent association between subtle morphological differences between morphs and their use of trophic resources. Skeletal anomalies and meristic counts did not corroborate the hypothesis that morphometric divergence in reversed individuals may be caused or reinforced by developmental instability. J. Morphol. 2012.

Use of neural network models to predict diatom assemblages in the Loire-Bretagne basin (France)

The aim of our work was to test the accuracy of the artificial neural networks (ANN) as predictive tools for benthic diatom taxa presence starting from a set of environmental variables. The river basin we studied is characterized by a huge complexity both in terms of spatial heterogeneity, as far as the environmental information is concerned, and in terms of biotic information, because of the large number of taxa that have been identified. In particular, this study focused on the application of different approaches to the reduction of the complexity of the data set and of the ANN models. In the meantime, it was also aimed at showing, as already pointed out for other kind of organisms (Scardi et. al. in # 3.8, Di Dato et al. in # 4.3), that too frequent or too rare species usually provide trivial information about the relationships of environmental variables with their presence or absence, thus affecting the accuracy of the models. The taxa that were selected according to the different approaches we tried including only those that can be actually modelled on the basis of the available environmental information. The limits and the perspectives of these species selection approaches are then thoroughly discussed.