Unsupervised functional classification applied on high resolution oceanographic data in Canaries Current Large Marine Ecosystem : toward fine scale analysis
Résumé
The understand of the fine scale process occurring in the ocean needs high resolution data and ad hoc analysis approach to improve the knowledge of ecosystem functioning. During an international survey carry out in 2014'AWA’ on-board the research vessel Thalassa (Ifremer, Brest) along the coast of Mauritania and Senegal we have used simultaneously multifrequency scientific echosounder and a Scanfish, both system allow a continuous acquisition of high- quality data at high spatial and temporal resolution over long distance. The functional data analyses have recently raising in serval field of statistics and appear to be well suited for the analysis of this dataset. In fact such data has spatial-functional nature and may be considered as observations of a stochastic process X in space of continuous functions over an interval T. Let X1(t), X2(t),.., Xn(t), t T, be the collection of n observations from X. First, we study an eventual horizontally or vertically variation of the acoustic intensity, we consider for a given frequency (here 200 kHz) and one vessel radial the two cases: vertical and horizontal variations of the acoustic intensity. Unsupervised functional classification used, shows a horizontal and vertical variation of acoustic intensity for a given frequency and a given radial. The approach can led to scrutinized at fine scale the processes occurring in three dimensions in the pelagic environment. The statistical functional classification applied to this case study appears powerful, ad hoc for ecological studies of marine ecosystem and will be extend to model the spatial structuration of the pelagic ecosystem according to the physcio-chemical parameters of the water mass which will allow to improve the forecast of the effect of the environment on marine ecosystem organization.