DART_LAB Tutorial

Overview

DART_LAB is a MATLAB®-based tutorial to demonstrate the principles of ensemble data assimilation. DART_LAB consists of PDF tutorial materials and MATLAB® exercises. See below for links to the PDF files and a list of the corresponding MATLAB scripts. The DART_LAB tutorial begins at a more introductory level than the materials in the tutorial directory, and includes hands-on exercises at several points.


Section 1The basics in 1D.
Section 2How should observations of a state variable impact an unobserved state variable? Multivariate assimilation.
Section 3Sampling error and localization.
Section 4The Ensemble Kalman Filter (Perturbed Observations).
Section 5Adaptive Inflation.


MATLAB® Hands-On Exercises

In the matlab subdirectory are a set of MATLAB scripts and GUI (graphical user interface) programs which are exercises that go with the tutorial. Each is interactive with settings that can be changed and rerun to explore various options. A valid MATLAB license is needed to run these scripts.

The exercises use the following functions:

function description
gaussian_product graphical representation of the product of two gaussians
oned_ensemble explore the details of ensemble data assimilation for a scalar
oned_model simple ensemble data assimilation example
oned_model_inf simple ensemble data assimilation example with inflation
run_lorenz_63 ensemble DA with the 3-variable Lorenz ‘63 dynamical model - the “butterfly” model
run_lorenz_96 ensemble DA with the 40-variable Lorenz ‘96 dynamical model
run_lorenz_96_inf ensemble DA with the 40-variable Lorenz ‘96 dynamical model with inflation
twod_ensemble demonstrates the impact of observations on unobserved state variables

To run these, cd into the DART_LAB/matlab directory, start matlab, and type the names at the prompt.