The folder coldpooling contains the python programs used to process automatic weather station output (simulated or not) for keras/tensorflow, to detect cold pools (CPs) in the data, and to train several kinds of neural networks, including one-shot and autoregressive LSTMs, to predict CPs in Dakar from time series. Jupyter notebooks are included, which were used to train and test the LSTM models.