SRGAN with Multi-Task Approach for Climate Downscaling

https://medium.com/@jackswl/climate-downscaling-using-srgan-with-multi-task-learning-731c8257ef4d

Acknowledgement: This article is an excerpt of my project during NUS Undergraduate Research Opportunity Programme, where I worked with Prof. Xiaogang HE, Zhanwei LIU and Huimin WANG. Without them, the idea, research and codes provided in this article would not have been possible.

Check out Zhanwei LIU's latest paper: https://www.nature.com/articles/s44221-023-00126-0

model file contains the model architecture notebooks file contains the files required to classify and downscale GEFSv12 reanalysis data, and preprocessing of GEFS ensemble input postprocessed_data contains all the GEFS ensemble (input), y_class (ground truth) and y_hr (ground truth) files

Step 1: Obtain y_hr_train.npy, y_hr_val.npy, and y_hr_test.npy by running 'downscaling_GEFS_reanalysis.ipynb' in 'notebooks' folder

Step 2: Run model folder with all the training and validating postprocessed .npy files

Step 3: Obtain the best .h5 models to test on the test .npy files