The main objective of this thesis is the characterization of high frequency dynamics in coastal areas and in particular their extremes, through the study of long-term biodeochemical time series registered by automated systems. The majority of high-frequency data sets used in this study came from MAREL program. The low-frequency time series from coastal monitoring programs SOMLIT (CNRS, INSU) and SRN (Ifremer) are employed to support the importance of automated systems. The EMD (Empirical Mode decomposition) method has provided a basis for us to study several of these time series. We also have used some methods more classical borrowed from numerical analysis field and turbulence. This study is organized in three chapters, and several appendices. The first chapter is devoted to the material and method. In the second chapter, using the EMD method we have highlighted the strong fluctuations contained in the blooms, and we have performed spectral analyzes. The principal component analysis (PCA) highlighted the main forcing exerted on primary production and SOMLIT temperature profiles suggest an impact of stratification on the intensity of blooms. In the third chapter, we conducted a comparative study between low-frequency and high-frequency data. Two cross-correlation methods (TDIC and co-spectra) allowed us to define a characteristic transition scale between the temperatures of the western and eastern English Cahnnel. In appendices we tested the robustness of different spectral analysis methods about the missing data in the time series, which is an underlying problem in the database registered by automated systems, and we reproduce a paper, which is under submission.
展开▼
