Generate Taylor Diagram from the output metrics of corrected data to the source

taylor_diagram_outmetrics.py

# -*- coding: utf-8 -*-
"""
NAME
    taylor_diagram_outmetrics.py
    Generate Taylor Diagram from all metrics output
DESCRIPTION
    Input data for this script will use metric.csv generated by imerg_cpc_biascorrection.py.
REQUIREMENT
    It required numpy, pandas, metpy, matplotlib and xarray module. So it will work on any machine environment
EXAMPLES
    python taylor_diagram_outmetrics.py
NOTES
    To visualize the metrics output using a Taylor diagram, you can use the taylor_diagram function 
    from the taylor_diagram module of the metpy library. This function allows you to plot the relative 
    bias, Pearson correlation coefficient, and root mean square error of a dataset on a Taylor diagram, 
    with reference to a reference dataset.
CONTACT
    Benny Istanto
    Climate Geographer
    GOST, The World Bank
LICENSE
    This script is in the public domain, free from copyrights or restrictions.
VERSION
    $Id$
TODO
    xx
"""
import os
import xarray as xr
import pandas as pd
import numpy as np
from metpy.plots import taylor_diagram
import matplotlib.pyplot as plt

# Define the list of methods
methods = ['scale', 'distribution', 'delta', 'lsc', 'lscdf', 'rcdfm', 'mlr', 
'ann', 'kalman', 'bcsd', 'bcsdm', 'qq_mapping', 'eQM', 'aQM', 'gQM', 'gpdQM']

# Define the appropriate input and output directory paths
output_dir = f'output'
diagram_dir = f'{output_dir}/diagram'

# Create the output directories if they don't already exist
os.makedirs(output_dir, exist_ok=True)
os.makedirs(diagram_dir, exist_ok=True)

for method in methods:
    # Load the metrics output from the bias correction
    metrics = pd.read_csv(f'output/{method}/metrics/{method}_metrics.csv')

    # Select the reference dataset (CPC in this example) and the corrected datasets (IMERG ann in this example)
    ref = metrics[metrics['dataset'] == 'ref']
    corr = metrics[metrics['dataset'] == method]

    # Extract the relative bias, Pearson correlation coefficient, and root mean square error for each dataset
    ref_r, ref_p, ref_rmse = ref['relative_bias'], ref['pearson'], ref['rmse']
    corr_r, corr_p, corr_rmse = corr['relative_bias'], corr['pearson'], corr['rmse']

    # Create the Taylor diagram figure
    fig = plt.figure(figsize=(10, 10))
    ax = fig.add_subplot(111, projection='polar')

    # Add the reference dataset to the Taylor diagram
    taylor_diagram(ref_rmse, ref_r, ref_p, marker='o', color='k', markersize=10, label='ref')

    # Add the corrected dataset to the Taylor diagram
    taylor_diagram(corr_rmse, corr_r, corr_p, marker='s', color='b', markersize=7, label=f'{method}')

    # Add the gridlines and legend to the Taylor diagram
    plt.grid(True)
    ax.legend(loc='upper left')

    # Save the Taylor diagram to a PNG file in the output folder
    plt.savefig(f'{diagram_dir}/{method}_taylor_diagram.png', dpi=300, bbox_inches='tight')

    # Show the Taylor diagram
    plt.show()