Encoder-Decoder-Model-for-Rainfall-Runoff-Modeling (Case of Upper Awash Basin in Ethiopia)

Implementation of different LSTM models including Encoder-Decoder based approaches for Rainfall-Runoff Modeling for Awash River in Ethiopia. Time series prediction is a widespread problem. Applications range from price and weather/flood/rain-fall runoff forecasting to biological signal prediction.

In this project I try to demonstrate implemention of a Recurrent Neural Network (RNN) based vanilla LSTM/GRU and encoder-decoder models for time series prediction task using Keras. I have used a case study of the Upper Awash basin Meteorological and Hydrological data which is recorded at various record stations. I have used 28 years of daily recorded data.

In this repo I provide the data used and the detailed experimental setups for the different vanilla LSTM/GRU and Encoder-Decoder based approaches. You can download the zip file of this ripo and run it on Google Colab. Details of the experiments are found on the provided notebook files.

Contents of this repository

• Vanilla LSTM/GRU Code -- A folder containing Jupyter notebooks for vanilla LSTM/GRU experiments
• Encoder-Decoder Code -- A folder contatining Jupyter notebooks for encoder decoder based experiments
• 1981_final -- Folder containing all the datasets used in this project

Contact

Fetulhak Abdurahman: afetulhak (at) yahoo.com