Evaluate potential correlations between the occurrence of extreme winter warming and some types of Arctic vegetation dying (in particular mosses and lichens).

This is what the various Jupyter notebooks do:

prob_mean_tp1n.ipynb the inputs and outputs for the ML algorithm :

• taking Copernicus World Land Cover data at 100m x 100m resolution from 2015 to 2019
• identifying each year the locations with moss & lichen
• extracting the corresponding ERA5-Land 2m temperature (t2m) and total precipitation (tp)
• also finding WLC data for the following year

and produces .hdf files with x_year and y_year);

merge_mean_tp1n.ipynb combines the yearly .hdf files into input & output .csv files;

deep_mean_tp1n.ipynb reads the input and output .csv files and split them into X_train, X_test, y_train and y_test (80% for training and 20% for testing, randomly shuffled);

train_mooc_tp1n.ipynb

• instantiates a keras.Input class
• defines the hidden layer and the corresponding activation function
• creates the output layer with the output activation
• creates a Keras model
• compile the model with a Huber loss function and the Adadelta optimizer
• trains the model a number of epocs
• plots the loss history
• performs a forecast.

CGLC_download.ipynb

• downloads Coperncus Global Land Cover data (100x100m horizontal resolution)

ERA5-land_download.ipynb

• downloads ERA5-land data hourly, from 2015

(Note that writing to CESNET' s3 requires credentials bosdisclosed i the notebooks)