# Acquisition et analyse des données, compléments

(Difference between revisions)

## Revision as of 09:47, 5 March 2012

Materials for Data acquisition and analysis (OCEA0035-1)

## Organisation

This lecture is given the first semester each year. We will review theoretical concepts needed for the course, but mainly we will focus on the application of various data analysis techniques to several data sets. The exercises are done in Matlab or Octave and the necessary code will be either provided or developed during the lecture.

# Exercises

## Exercise 1

Quality control: using file 8762075.sealevel.txt, representing the hourly sea level height on the west Florida Shelf in 2004, detect the suspect data that might be classified as outliers. Discussion on why these suspect data should/should not be classified as outliers.

## Exercise 2

Exploratory data analysis

From the World Ocean Dataset:

• Read the data and transform them to Ascii (program wodASC_all.f). Extract at least temperature and salinity.
• Reading in matlab and exploratory representation of data and analyses:
• Detection and elimination of outliers (minimum and maximum threshold, deviation from mean...)
• Other suspect data?
• Representation of data:
• Seasonal scatter plots (different seasons/months, different depths...)
• Time series (spatial averages)
• Vertical profiles (different seasons, geographical zones)
• Distribution of variables
• T/S diagrams, Hovmöller diagrams...
• Discussion on representation of data: scales, readability of axes, units ...

## Exercise 3

Linear regression

To illustrate the application of linear regression and calculation of trends, atmospheric carbon dioxide from Mauna Loa (Hawaii) will be used.

Data: CO2 (ppmv) and time dimension

Apply a linear regression to the data (code here). Discuss if a linear regression is appropriate in this case. Calculate a linear regression by periods, and compare the trends for each period.

For the final work, a linear regression will be applied to the WOD dataset of Exercise 2. Is there a trend in your data? Does it change with time? Compare the trend at different depths.

## Exercise 4

Filters and interpolation

Using a time series from the data in exercise 2 (a spatial average, for example), apply a Gaussian-window filter to extract the annual cycle. If there are small gaps in the time series, an interpolation of the data to a regular time step will be done. Compare between the Linear and spline methods of interpolation.

## Exercise 5

Fourier decomposition

Apply a Fourier decomposition to the data in exercise 1 and detect which are the frequencies that dominate the time series. To which phenomena do these correspond?

For the final work, apply a Fourier decomposition to a time series of your data (must be gap-free) and comment on the frequencies that dominate them. Explanation of phenomena that may give rise to these frequency peaks.

## Exercise 6

Error Assessment

Using these data from the Cariaco basin (Venezuela), assess the error of the model respect to the observations using the various error measures viewed during the lecture. Explanation of what do we learn from these results about the model performance.

For the final work, include this discussion on model errors. Try to infer a physical reason for the presence of errors. If you have model (and data) you would like to use, then use them instead of the example data.

## Exercise 7

EOF decomposition

We will use sea surface height (SSH) data from the Caribbean Sea (here is the latitude, the longitude and the time). Perform an EOF decomposition on these SSH data. Comment at least the 3 first EOFs. What is represented at each EOF? What are the spatial and temporal scales? What is the percentage of variability of each EOF mode?.

The presentation given during the lecture is here.