Statistics Short Course, AOS Princeton

Statistics for Atmospheric and Oceanic Sciences

18 July - 28 July 2006

Correlation between the Southern Annular Mode (SAM) index and wind speed. Image courtesy of Nicole Lovenduski, Department of Atmospheric and Oceanic Sciences, The University of California, Los Angeles 2006

Course Description

This is a graduate level short course in statistical methods frequently used to interpret model results and observations in earth sciences. Each session will consist of two parts: a one hour lecture followed by an approximately 2 hour practical in which students will have the opportunity to apply the methods learned to atmospheric and oceanic data and models.

All classes will be held in 312 Sayre Hall, Mondays, Wednesdays, and Fridays from 1-4 PM. The first class (Monday July 17) had been moved to Tuesday July 18 due to a conflict with John Dunne's seminar at GFDL.

Materials for the Practical Sessions

Since there is no computer lab available for our practical exercises, it would be greatly appreciated if all students that own laptops would bring them. Please check this space before each class to download the materials that will be needed for the class. These exercises can be done in pairs or small groups, so please don't worry if you don't have your own laptop.

Computing and Visualization Software

You are free to use any software that they like for community exercises, however you will need to have something appropriate for calculations, matrix manipulations, and will need to be able to read and write netCDF files.

If you will be using Matlab, and do not already have a favorite method for reading and writing netCDF files, Sara recommends the free CSIRO netCDF interface: marine.csiro.au/sw/matlab-netcdf.html. Installation instructions are in the introduction. You will also need Chuck Denham's netCDF toolbox, mexcdf.sourceforge.net/netcdf_toolbox.html. Please note that you need to add the appropriate paths before accessing these toolboxes in Matlab.

Andy Jacobson will be using R in his lectures, which is a language environment designed for statistical computing. It is free, and can be downloaded from www.r-project.org. Students are not required to use this software, but Andy says there will be "greatly enhanced benefits" for using R. If you will be using R, please install it prior to the first class along with the following packages from cran.r-project.org: ncdf, fields, and e1071. This can be done by starting R and typing:

install.packages(c("ncdf","fields","e1071"))

Additional R resources: click here. Questions and comments about R can be directed to Andy: andy.jacobson@noaa.gov.

Ferret is also highly recommended for visualization; however, it may not be the best choice for some of the statistical computations that will be done in this class. Ferret is free and can be downloaded from ferret.wrc.noaa.gov/Ferret/.

Materials for 28 July 2006

You will need the following netCDF files containing precipitation from slightly different versions of CM2.1.

cm2.1U_p01.1-100.precip.nc
cm2.1U_p002.1-100.precip.nc

In addition, for some optional further exploration of skewness, you are welcome to use:
Satellite SST Data
The first 100 years of output from the CM2.1 control run.

Lectures

Lecture slides and other materials will be posted here after each class

18 July 2006- Correlations and significance, Sara Mikaloff Fletcher
lecture slides (pdf)
practical (pdf)

Follow up materials
For a little help with Matlab, Ferret, and netCDF files, check out the following:
A (very) short cheat sheetof handy Matlab commands
A (very) short cheat sheetof handy Ferret commands
An example of a Matlab routine to write a netCDF file.

For some nice examples of how correlations in observed quantities can be used to learn about physical processes, you can have a look at the following papers:
Lovenduski et al., 2005 (pdf) (At least one figure in this paper should look a little familiar.)
Wilson et al., 2001 (pdf)

19 July 2006- Linear regressions, Andy Jacobson
lecture slides (pdf)

21 July 2006- EOF, Andy Jacobson
lecture slides (pdf)

24 July 2006- Uncertainty estimates from error propagation, Monte Carlo,and Boot Strap methods, Sara Mikaloff Fletcher lecture slides (pdf)
practical (pdf)

Follow up materials
For further reading about the gas exchange tracer used in today's exercises you might enjoy Gruber and Sarmiento, 2002.

26 July 2006- Evaluating models: Skill scores, and Taylor diagrams, Sara Mikaloff Fletcher
lecture slides (pdf)
practical (pdf)

Follow up materials
Jim Orr and Patrick Brockmann of LSCE, France have provided scripts to generate Taylor diagrams using Ferret on the Ferret Web Site. In addition, Jean Marie Epitalon has provided a program to do it in Python on his web site.

For further reading about Taylor diagrams, you might enjoy Taylor, 2001.

If you would like to see some further examples of Taylor diagrams in action, check out the OCMIP web site or Raynaud et al., 2006

28 July 2006- Model differences, Skewness, and Symmetry, Sara Mikaloff Fletcher
lecture slides (pdf)
practical (pdf)

Follow up materials
For further reading about skewness and kurtosis and ENSO, you might enjoy Burgers and Stephenson, 1999.

Instructor Contact Information

Sara Mikaloff Fletcher

Atmospheric and Oceanic Sciences
Sayre Hall, Forrestal Campus
Princeton University
Princeton, NJ 08544

tel: 609-258-8340
sara@ splash.princeton.edu

Andrew Jacobson

NOAA Earth System Research Lab
Global Monitoring Division
325 Broadway
Boulder, CO 80305

tel: 303-497-4916
andy.jacobson@noaa.gov