Plot Exceedance Frequency Curve

[matamadio]

Use the results to plot a chart of Annual Exceedence Probability and related EAI at the end, above or below map:

• The total exposure (blue chart) has RP frequency as X, and total exposure as Y.
• The total impact (orange chart) has RP frequency as X, and impacted exp as Y; and a label for total EAI.

As in:

[matamadio]

Update on previous with better EAI calculation

The calculation performed in the table depends on the number (var) of RPs used. In the analysis:

Exp_EAI: Sum(RPi_exp_imp*(1/RPi - 1/RPi+1)

e.g. with RP selection [5, 10, 25, 50]:
RPi = 5
RPi+1 = 10

See example for Exp=Pop

	RP5_pop_imp	RP10_pop_imp	RP20_pop_imp	RP50_pop_imp	RP75_pop_imp	RP100_pop_imp	RP200_pop_imp	RP250_pop_imp	RP500_pop_imp	RP1000_pop_imp	POP_EAI
Impact	0.87	65.575	67.475	68.881	69.497	70.07	70.727	71.058	72.037	72.869
EAI	0.09	3.28	2.02	0.46	0.23	0.35	0.07	0.14	0.07	0.07	6.79
Calc:	B2*(1/5-1/10)	C2*(1/10-1/20)	D2*(1/20-1/50)	E2*(1/50-1/75)	F2*(1/75-1/100)	G2*(1/100-1/200)	H2*(1/200-1/250)	I2*(1/250-1/500)	J2*(1/500-1/1000)	K2*(1/1000)

For each ADM, we are interested to keep in the final table only the columns:

• RPi_exp_imp
• Exp_EAI (POP_EAI in the calc example).

Note that the EAI value changes with the number of RP selected, where the most reliable is 10; a selection of only 3 RP leads to results with a ratio of 56% compared to 10RP EAI in this example over DOM.

[artessen]

Here, a first version of a Python notbook to create the plot function for EAI using matplotlib and seaborn
Sample file to run the code here:
NGA_RSKv2.csv

# Importing the required libraries
import datetime, os, shutil, gc, socket
import math
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from matplotlib.ticker import FormatStrFormatter, StrMethodFormatter
import seaborn as sns

# Reading the data for plotting
dfData = pd.read_csv('NGA_RSKv2.csv') 

# Defining the title and sub-title
title    = 'Flood x Population EAI'
subtitle = 'Exceedance frequency curve'

# Defining the x- and y-axis data and text
x  = dfData[['Freq']].values.squeeze()
y  = dfData[['Pop_impact']].values.squeeze()
xlbl = 'Frequency'
ylbl = 'Impact (population)'

# Defining if plotting total EAI
txtTotal = True
xpos = 0.70
totEAI = np.sum(dfData[['Pop_EAI']].values.squeeze())

# Defining the plot style
sns.set_style('whitegrid')
blue, = sns.color_palette('muted', 1)
xFreqAsRP = True
yOnRight = False

# Getting the labels for RPs
rp_lbl = dfData[['RP']].values.squeeze().tolist()
rp_lbl = ['RP '+str(item) for item in rp_lbl]
x_offset = -0.003
y_offset = 20000
if yOnRight: x_offset = -1*x_offset
 
# Creating the plot
fig, ax = plt.subplots()
ax.plot(x, y, color=blue, lw=1, marker="o")
if xFreqAsRP:
    for i in range(len(rp_lbl)):
        ax.text(x[i]+x_offset, y[i]+y_offset, rp_lbl[i], color='#4D4D4D',
                bbox=dict(facecolor='#FAFAFA', edgecolor='#4D4D4D', boxstyle='round,pad=0.1', alpha=0.4))
ax.fill_between(x, 0, y, alpha=.3)
ax.set(xlim=(0, max(x)), ylim=(0, max(y)*1.05), xticks=np.linspace(0, max(x), 6))
ax.ticklabel_format(style='plain', axis='y')
ax.yaxis.set_major_formatter(StrMethodFormatter('{x:,.0f}'))
ax.set_xlabel(xlbl)
ax.set_ylabel(ylbl)
ax.set_title(str(title+'\n'+subtitle))
if yOnRight:
    ax.invert_xaxis()
    ax.yaxis.tick_right()
    ax.yaxis.set_label_position('right')
if txtTotal:
    vtext = 'EAI = '+'{:,}'.format(totEAI)
    if yOnRight: xpos = 0.04
    ax.text(xpos, 0.9, vtext, fontsize=14, color='black', ha='left', va='bottom', transform=plt.gca().transAxes, 
            bbox=dict(facecolor='#FAFAFA', edgecolor='#4D4D4D', boxstyle='square,pad=0.4', alpha=0.6));

# Show the graph
plt.show()

[matamadio]

Great!

Added it here: https://github.com/GFDRR/CCDR-tools/blob/main/Top-down/plots/Exceedance%20Frequency%20Curve.py

Would be nice to integrate directly in the script or results dasbhoard.

[matamadio]

@artessen
please check latest EAI calculation in the code; in the KHM output, it seems to be simply sum(RPi_imp/RPi) as in the first post; while it should use exceedence frequency instead as explained in updated approach.
I.e. RP5_EAI = RP5_IMP * (1/5-1/10)

[matamadio]

To add script in the repo, and possibly integrate in parallel code output @artessen