Data-Driven-Attribution-of-Climate-Events

---

1st Step: Reproduce the Paper from Badr et al (2014) 'Application of Statistical Models of the Prediction of Seasonal Rainfall Anomalies over the Sahel'

Note: The underlying data used here is not always the same as in Badr et al. (2014).
In general I used the most recent versions of the single data sets.

• nb_badr01_sahelrainfall.ipynb plots a time series from the Sahel Precipitation Index (SPI).

• nb_badr02_boxcorr.ipynb box-correlates the SPI with global Surface Temperature.

• nb_badr03_indices.ipynb reproduces Badr's predicor indices

• nb_badr04_pca.ipynb applies PC decomposition to the whole set of predictor variables

• nb_badr05_pred_nn.ipynb sets up and run a simple Neural Network for summer rainfall prediction

• nb_badr05_pred_nn_gridsearch_tf using tensorflow to perform Grid Search for optimizing Badr's model

• nb_ersst_data.ipynb merges the single ERSST Data files into one netCDF file.

• For the model prediction Badr et al. (2014) used a total set of 20 predictor variables.
  In nb_predictor_indices.ipynb the according 20 AMJ predictors are computed.

2nd Step: Modify model - Aiming for better performance

• nb_klausmodel_gridsearch apply more recent methods in model architechture