Convolutional GRU Network for the El Nino-Southern Oscillation Region Forecast

1. Create an virtual environment using requirements.txt

conda install --yes --file requirements.txt

2. Running details for the CCSM4 dataset:

2.1 Get the data. This is a preprocessed dataset generously provided by Prof. Dimitrios Giannakis. It is not contained due to size issue, and requests for CCSM4 dataset should be addressed to lingdaw2@illinois.edu

2.2 Train the ConvGRU network under the dir ConvGRU_CCSM4, and the trained models will be saved in ConvGRU_CCSM4/results/saved_ConvGRU_model

CUDA_VISIBLE_DEVICES=0 python main.py

2.3 Switch to the dir ConvGRU_CCSM4/results, and test the trained model

CUDA_VISIBLE_DEVICES=0 python ConvGRU.py

2.4 Get results for other methods

python GRU.py
python LIM.py
python LR.py

2.5 Switch to the dir ConvGRU_CCSM4/results/plots and use plot_res.ipynb for figures

3. Running details for the NOAA-GDFL-SPEAR dataset:

3.1 Download the data to the dir ConvGRU_SPEAR/data, and instructions can be found at ConvGRU_SPEAR/data/README.md

3.2 Train the ConvGRU network under the dir ConvGRU_SPEAR, and the trained models will be saved in ConvGRU_CCSM4/results/saved_ConvGRU_model

CUDA_VISIBLE_DEVICES=0 python main.py

3.3 Switch to the dir ConvGRU_SPEAR/results, and test the trained model

CUDA_VISIBLE_DEVICES=0 python ConvGRU.py

3.4 Get results for other methods

python GRU.py
python LIM.py
python LR.py

3.5 Use plot_res.ipynb for figures

4. Running details for the NOAA-CIRES dataset:

4.1 Download the data to the dir AirTmp_M/data, and instructions can be found at AirTmp_M/data/README.md

4.2 Train the ConvGRU network under the dir AirTmp_M, and the trained models will be saved in AirTmp_M/saved_ConvGRU_model

CUDA_VISIBLE_DEVICES=0 python main.py

4.3 Test the trained model

CUDA_VISIBLE_DEVICES=0 python plot_res.py

4.4 Switch to AirTmp_M/plots, and use plot_res.ipynb for figures