STConvS2S: Spatiotemporal Convolutional Sequence to Sequence Network for weather forecasting

This repository has the open source implementation of a new architecture termed STConvS2S. To sum up, our approach (STConvS2S) uses only convolutional neural network (CNN) to tackle the sequence-to-sequence task using spatiotemporal data. We compare our results with ConvLSTM architecture. Details in the paper: arXiv:1912.00134.

Requirements

Mainly, our code uses Python 3.6 and PyTorch 1.0. See config/environment.yml for other requirements.

To install packages with the same version as we executed our experiments, run the code below:

cd config
./create-env.sh

Datasets

All datasets are publicly available at http://doi.org/10.5281/zenodo.3558773. Datasets must be placed in the data folder.

Note: In the 15-step ahead datasets, just for the convenience of having the input and targe tensors in the same file, we define them in the same shape (# of samples, 15, # of latitude, # of longitude, 1). However, in the code, we force the correct length of the input sequence in the tensor as shown here.

Experiments

Jupyter notebooks for the first sequence-to-sequence task (given the previous 5 grids, predict the next 5 grids) can be found in the experiments folder (see Table 1 in the paper).

Final experiments (see Table 2 in the paper) compare STConvS2S (our architecture) with ConvLSTM architecture and a baseline (ARIMA models). We evaluate the models in two horizons: 5 and 15-steps ahead. This task is performed using main.py (for deep learning models) and baseline.py (for ARIMA models). Results can be found in the output folder.

• /output/full-dataset (for deep learning models)

  • /checkpoints: pre-trained models that allow you to recreate the training configuration (weights, loss, optimizer).
  • /losses: training and validation losses. Can be used to recreate the error analysis plot
  • /plots: error analysis plots from training phase
  • /results: evaluation phase results with metric value (RMSE or MAE), training time, best epochs and so on.

• /output/baseline: results of baseline models

Usage

First load the conda environment with the installed packages.

source activate pytorch

Below are examples of how to run each model.

STConvS2S

python main.py -i 10 -v 4 --plot > output/full-dataset/results/cfsr-stconvs2s-rmse-v4.out

The above command executes STConvS2S in 10 iterations (-i), indicating the model version (-v), allowing the generation of plots in the training phase (--plot).

ConvLSTM

To run experiments with ConvLSTM architecture, add the --convlstm parameter to the above command.

Baseline

python baseline.py > output/baseline/cfsr-arima.out

Additional parameters

• add --chirps: change the dataset to rainfall (CHIRPS). Default dataset: temperature (CFSR).
• add --mae: change the metric to MAE. Default metric: RMSE.
• add -s 15: change the horizon. Default horizon: 5.
• add --email: send email at the end. To use this functionality, set your email in the file config/mail_config.ini.

Citation

@article{Nascimento2019,
	author    = {Rafaela C. Nascimento, Yania M. Souto, Eduardo Ogasawara, Fabio Porto and Eduardo Bezerra},
	title     = {{STConvS2S: Spatiotemporal Convolutional Sequence to Sequence Network for weather forecasting}},
	journal   = {arXiv:1912.00134},
	year      = {2019}
}

Contact

To give your opinion about this work, send an email to rafaela.nascimento@eic.cefet-rj.br.