Emulator

Image sources 1, 2, and 3.

Scripts

• Merge shapefiles into a single multi-polygon (create_merged_shapefile.ipynb).
• Crop emissions data to China shapefile training (emis_apply_mask_train.py) and test (emis_apply_mask_test.py) data.
• Design of Latin hypercube designs (latin_hypercube.ipynb).
• Output of Latin hypercube designs for training (latin_hypercube_inputs_training.csv) and test (latin_hypercube_inputs_test.csv) data.
• Automatic machine learning tool (TPOT) using genetic programming to optimise model pipeline on 50 random grid cells (tpot_optimiser.py), with outputs in the tpot_gridcells folder (tpot_emulator_pipeline_PM2_5_DRY_*.py).
• Configuration for TPOT based on Gaussian process regresor (config_regressor_gaussian.py).
• Emulator cross-validation and sensitivity analysis (emulator.ipynb). Interactively computed on a HPC using Dask and Jupyter Lab following instructions here.
• Emulator predictions for custom inputs (emulator_predictions.py). Submitted to HPC (emulator_predictions.bash) using Dask for workers viewing worker status on Jupyter Lab. Can submit in batch mode (emulator_predictions_batch.bash).
• Regrid custom outputs to population grid of the world (regrid_to_popgrid.py). Submitted to HPC (regrid_to_popgrid.bash) using Dask for workers viewing worker status on Jupyter Lab. Can submit in batch mode (regrid_to_popgrid_batch.bash).
• Crop population-weighted output predictions to region's shapefile (popweighted_region.py). Submitted to HPC (popweighted_region.bash) using Dask for workers viewing worker status on Jupyter Lab. Uses cropping functions (cutshapefile.py).
• Various emulator plots including emulator evaluation, sensitivity maps, prediction maps, and 2D contour pairs, (emulator_plots.ipynb).
• Plots of the emissions (emission_plots.ipynb).

Setup Python environment

• Create a conda environment with the required libraries from the config file (.yml) in the repository:

conda env create --name pangeo --file=pangeo.yml  
pip install salib dask_labextension pyarrow  
jupyter labextension install dask-labextension  
jupyter labextension install @jupyter-widgets/jupyterlab-manager  

License

This code is currently licensed under the GPLv3 License, free of charge for non-commercial use.