This Julia package provides tools to train Echo State Networks (ESNs) to predict climate indices. It implements two different methods for TIme-Delay Embeddings (TDEs). We compare the prediction performance of ESNs trained with and without those TDEs on the El Niño - Southern Oscillation and the Madden - Julian Oscillation. Therefore, we measure the prediction accuracy of ESNs across lead-times, employing metrics like the Root-Mean-Square-Error (RMSE), Pearson Correlation Coefficient (PCC), or the Heidke-Skill-Score (HSS).
