Skip to content

Commit

Permalink
make obsrevation dir
Browse files Browse the repository at this point in the history
  • Loading branch information
@ERKuipers
ERKuipers committed Nov 26, 2024
1 parent 2a9b635 commit 0b55072
Showing 1 changed file with 31 additions and 26 deletions.
57 changes: 31 additions & 26 deletions GloFAS/GloFAS_analysis/performance_calculator.py
View file
Original file line number Diff line number Diff line change
@@ -1,11 +1,11 @@
import pandas as pd
import geopandas as gpd
import os
import numpy as np
import unidecode
import matplotlib.pyplot as plt
from datetime import datetime, timedelta
from GloFAS.GloFAS_analysis.flood_definer import FloodDefiner

from GloFAS.GloFAS_prep.vectorCheck import checkVectorFormat
import GloFAS.GloFAS_prep.configuration as cfg
class PredictedToImpactPerformanceAnalyzer:
Expand All @@ -31,7 +31,8 @@ def __init__(self, DataDir, RPyr, leadtime, impactData, triggerProb, adminLevel,
self.impactData = impactData
self.PredictedEvents_gdf = PredictedEvents_gdf
self.comparisonType = comparisonType

self.comparisonDir = f'{self.DataDir}/{comparisonType}'
os.mkdir (self.comparisonDir, exist_ok=True)
if comparisonType =='Observation':
self.impact_gdf = self.openObservation_gdf()
elif comparisonType =='Impact':
Expand Down Expand Up @@ -144,25 +145,25 @@ def clean_and_add_GloFAS (self, PredictedEvents_gdf):
# Keep entries from the original GloFASCommune_gdf where there was not yet a match
PredictedEvents_gdf = PredictedEvents_gdf[PredictedEvents_gdf['Remove'] != 1] # remove was a 1, so we must keep everything that is not 1

# NO YEAR IMPACT? --> no checking : this is the piece of code:
# for year in self.years:
# # Convert year to date range for that year
# start_of_year = pd.to_datetime(f'{year}-01-01')
# end_of_year = pd.to_datetime(f'{year}-12-31')
#NO YEAR IMPACT? --> no checking : this is the piece of code:
for year in self.years:
# Convert year to date range for that year
start_of_year = pd.to_datetime(f'{year}-01-01')
end_of_year = pd.to_datetime(f'{year}-12-31')

# # Find communes that recorded an impact in the given year
# recorded_impacts = self.impact_gdf[
# (self.impact_gdf['Start Date'] >= start_of_year) &
# (self.impact_gdf['End Date'] <= end_of_year)
# ][f'ADM{self.adminLevel}'].unique()
# Find communes that recorded an impact in the given year
recorded_impacts = self.impact_gdf[
(self.impact_gdf['Start Date'] >= start_of_year) &
(self.impact_gdf['End Date'] <= end_of_year)
][f'ADM{self.adminLevel}'].unique()

# # Remove rows in GloFAS where no impact was recorded for that commune in that year
# PredictedEvents_gdf = PredictedEvents_gdf[
# ~(
# (PredictedEvents_gdf['StartValidTime'] >= start_of_year) &
# (PredictedEvents_gdf['EndValidTime'] <= end_of_year) &
# (~PredictedEvents_gdf[f'ADM{self.adminLevel}'].isin(recorded_impacts))
# )]
# Remove rows in GloFAS where no impact was recorded for that commune in that year
PredictedEvents_gdf = PredictedEvents_gdf[
~(
(PredictedEvents_gdf['StartValidTime'] >= start_of_year) &
(PredictedEvents_gdf['EndValidTime'] <= end_of_year) &
(~PredictedEvents_gdf[f'ADM{self.adminLevel}'].isin(recorded_impacts))
)]

# Prepare the remaining rows to be added to impact_gdf
remaining_rows = PredictedEvents_gdf.copy()
Expand Down Expand Up @@ -207,7 +208,7 @@ def matchImpact_and_Trigger(self):
# now add remaining GloFAS rows
# add rows for trigger entries = 1 , but no recorded impact

self.impact_gdf.to_file (f"{self.DataDir}/{comparisonType}/model_vs{comparisonType}RP{self.RPyr:.1f}_yr_leadtime{self.leadtime:.0f}.shp")
self.impact_gdf.to_file (f"{self.DataDir}/{comparisonType}/model_vs{comparisonType}RP{self.RPyr:.1f}_yr_leadtime{self.leadtime:.0f}.gpkg")

def calc_performance_scores(self, obs, pred):
'''Calculate performance scores based on observed and predicted values,
Expand All @@ -218,16 +219,19 @@ def calc_performance_scores(self, obs, pred):
hits = df[['cons_class', 'hits']].drop_duplicates().hits[df.hits == True].count()
false_alarms = (pred.loc[pred.shift() != pred].sum()) - hits
misses = sum((obs == 1) & (pred == 0))
correct_negatives = sum((obs == 0) & (pred == 0))
correct_negatives = sum((obs == 0) & (pred == 0)) # there is no use in calculating correct negatives as there are many

# Calculate metrics
output = {
'pod': hits / (hits + misses) if hits + misses != 0 else np.nan, # Probability of Detection
'far': false_alarms / (hits + false_alarms) if hits + false_alarms != 0 else np.nan, # False Alarm Ratio
'pofd': false_alarms / (false_alarms + correct_negatives) if false_alarms + correct_negatives != 0 else np.nan, # Probability of False Detection
'csi': hits / (hits + false_alarms + misses) if hits + false_alarms + misses != 0 else np.nan, # Critical Success Index
'accuracy': (hits + correct_negatives) / (hits + correct_negatives + false_alarms + misses) if hits + correct_negatives + false_alarms + misses != 0 else np.nan, # Accuracy
'precision': hits / (hits + false_alarms) if hits + false_alarms != 0 else np.nan # Precision
# 'accuracy': (hits + correct_negatives) / (hits + correct_negatives + false_alarms + misses) if hits + correct_negatives + false_alarms + misses != 0 else np.nan, # Accuracy
'precision': hits / (hits + false_alarms) if hits + false_alarms != 0 else np.nan,
'TP': hits,
'TN': false_alarms,
'CN': correct_negatives
}

return pd.Series(output)
Expand All @@ -242,7 +246,8 @@ def calculateCommunePerformance(self):
lambda x: self.calc_performance_scores(x[f'{self.comparisonType}'], x['Event'])
)
scores_byCommune_gdf = self.gdf_shape.merge(scores_by_commune, on=f'ADM{cfg.adminLevel}')
scores_byCommune_gdf.to_file (f"{self.DataDir}/{comparisonType}/scores_byCommuneRP{self.RPyr:.1f}_yr_leadtime{self.leadtime:.0f}.shp")
scores_byCommune_gdf.to_file (f"{self.DataDir}/{comparisonType}/scores_byCommuneRP{self.RPyr:.1f}_yr_leadtime{self.leadtime:.0f}.gpkg")
scores_byCommune_gdf.drop(columns='geometry').to_csv (f"{self.DataDir}/{comparisonType}/scores_byCommuneRP{self.RPyr:.1f}_yr_leadtime{self.leadtime:.0f}.csv")
return scores_byCommune_gdf

if __name__=='__main__':
Expand All @@ -253,10 +258,10 @@ def calculateCommunePerformance(self):
for leadtime in cfg.leadtimes:
floodProbability_path = cfg.DataDir/ f"floodedRP{RPyr}yr_leadtime{leadtime}_ADM{cfg.adminLevel}.gpkg"
floodProbability_gdf = checkVectorFormat (floodProbability_path)
#calculate the flood events
# calculate the flood events
definer = FloodDefiner (cfg.adminLevel)
PredictedEvents_gdf = definer.EventMaker (floodProbability_gdf, cfg.actionLifetime, cfg.triggerProb)
#print (readcsv(f"{DataDir}/Données partagées - DNH Mali - 2019/Donne╠ües partage╠ües - DNH Mali - 2019/De╠übit du Niger a╠Ç Ansongo.csv"))
# print (readcsv(f"{DataDir}/Données partagées - DNH Mali - 2019/Donne╠ües partage╠ües - DNH Mali - 2019/De╠übit du Niger a╠Ç Ansongo.csv"))
analyzer = PredictedToImpactPerformanceAnalyzer(cfg.DataDir, RPyr, leadtime, hydro_impact_gdf, cfg.triggerProb, cfg.adminLevel, cfg.admPath, cfg.startYear, cfg.endYear, cfg.years, PredictedEvents_gdf, comparisonType)
analyzer.matchImpact_and_Trigger()
analyzer.calculateCommunePerformance()
Expand Down

0 comments on commit 0b55072

Please sign in to comment.