Analyse de données 2

Dernière mise à jour le 27 février 2024

Présentation

This course introduces “classic” elementary linear methods widely used for the analysis of insitu, satellite, numerical modelling data. Classes are hands-on sessions with computers during which students are taught to implement numerical algorithms in order to tackle specific problems of “fitting”, interpolation, principal component analysis. Some of these methods will be applied to real insitu physical data collected on board research vessel Albert Lucas during a day trip at sea in the bay of Brest.

Pré-requis nécessaires

elementary knowledge in linear algebra, scientific programming, python, statistics.

Objectifs

In addition to mastering the numerical implementation of the algorithm, students should understand the domain of application and the limitations of the fore-seen methods in order to re-use and adapt them for specific problems.

Compétences visées

Identify a data analysis method to tackle a specific problem
Build and adapt efficiently numerical algorithms designed for data analysis purposes
Validate the results obtained (error estimation)
Estimate the spatial and time interactions within a system using deterministic correlations

Descriptif

During this course, Students will be taught some key elementary linear methods used for the analysis of data either collected from insitu and satellite satellite observations, of produced by the integration of numerical models.Classes will be mostly composed of hand on sessions using Jupyter-Notebook, during which students implement numerical algorithms. The course is organized in 5 chapters:

linear regression and least square fitting
piecewise linear and cubic interpolation
optimal interpolation
2D fourrier spectrum
empirical orthogonal functions

Students will handle a written report of the analysis of the physical insitu data collected at sea.