This repository is a implementation of Climate Model Driven Seasonal Forecasting Approach with Deep Learning, accepted to Environmental Data Sciences Journal. For further inquiries about the code, please contact busraasan2@gmail.com.
To install requirements:
pip install -r requirements.txt
Even though CMIP6 and ERA5 is available online, our regridded data is not available yet. This section will be updated when it is ready.
In order to train the model, use the following command after configuring the args_parser.py file. In args_parser, one can set configurations such as what model, the number of lag years, predecessors/successors to use together with training parameters such as learning rate, weight decay etc.
python cimp6_trainer_.py
For the evaluation, the command below can be used. Same flags are available to training code is also available on this script.
python era_evaluate.py
In order to train models for monthly average output, command below can be used.
python cimp6_trainer_average.py
Comparison against CMIP6 predictions.
Experiments Benchmark
