This repository contains the source codes used in the following publications:

• Papagiannopoulou et al., 2017. "A non-linear Granger-causality framework to investigate climate–vegetation dynamics". Geoscientific Model Development. https://doi.org/10.5194/gmd-2016-266.
• Papagiannopoulou et al., 2017. "Vegetation anomalies caused by antecedent precipitation in most of the world". Environmental Research Letters. https://doi.org/10.1088/1748-9326/aa7145.

The script is mainly built using the following packages:

python 2.7.12 | Anaconda 2.3.0 (64-bit) (or higher)
scikit-learn

In order to test the framework, a test file is provided (test.csv). This file contains all features used in the publication for one pixel on earth.

The script allows you to test both the "linear" and "non-linear" framework. In order to execute these, use the following commands:

python GC_script.py test.csv linear
python GC_script.py test.csv non-linear

The outcome of both tests provide you information on the explained variance of the baseline and full model as well as a quantification of the Granger causality. For more information, see publication.

The output should look as follows:

Explained variance of baseline model: 0.076398
Explained variance of full model: 0.199819
Quantification of Granger causality: 0.123421
Total time: 58 seconds

Explained variance of baseline model: 0.055405
Explained variance of full model: 0.306413
Quantification of Granger causality: 0.251008
Total time: 6 seconds

This project is licensed under the GNU General Public License v3.0 - see the LICENSE file for details