- install CONDA virtual environment.
$ conda env create -f ./env/conda_env_venv2.yaml
$ conda env create -f ./env/conda_env_ncl.yaml
-
edit main script.
- For default run (single-input): edit main_run_def.bash
- For multi-input run: edit main_run_multi.bash
-
run main script.
$ ./main_run_def.bash # for default run
$ ./main_run_multi.bash # for multi-input run-
Files
- deep_model.py: contains deep-learning based rain rate nowcasting models
- deep_utils.py: contains utilities for data processing and loss functions, and etc.
- main_run_def.bash: main script for 'default (single-input)' run
- main_run_multi.bash: main script for 'multi-input' run
- sample_list_2012-2019_JJAS_RDR_avg1h_1hrs.csv: list for non-trivial rainfall cases during June-September, 2012-2019 over the Korean Peninsula
- setting.txt: templete file for recording the settings of each experiment
- step_00_fit_def.py: program for training a model using single-input
- step_00_fit_multi.py: program for training a model using multi-input
- step_01_pred_def.py: program for predicting a rain rate distribution using a model trained with single-input
- step_01_pred_multi.py: program for predicting a rain rate distribution using a model trained with multi-input
- step_02_draw_results.ncl: program that draws the predicted rain rate distributions through step_01
-
Directories
- ./env: contains conda environments
- ./data: contains preprocessed radar rainrate data
- ./output/ckpt: output directories for model checkpoint
- ./output/log: output directories for log (used by tensorboard)
- ./output/image: output directories for resulted images (outputs from 'step_02')
- ./output/script: output directories for used scripts
- ./output/predict: output directories for prediction (outputs from 'step01')