-
Notifications
You must be signed in to change notification settings - Fork 1
/
climatologies_step4_validation.py
317 lines (263 loc) · 16.4 KB
/
climatologies_step4_validation.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
#!/usr/bin/env python
# -*- coding: utf-8 -*-
__author__ = "Hylke E. Beck"
__email__ = "hylke.beck@gmail.com"
__date__ = "November 2022"
import os
import sys
import pdb
import time
import glob
import pandas as pd
import numpy as np
import tools
from datetime import datetime
from netCDF4 import Dataset
import warnings
import pickle
import matplotlib as plt
def main():
#==============================================================================
# Settings
#==============================================================================
warnings.filterwarnings('ignore')
np.set_printoptions(suppress=True)
config = tools.load_config(sys.argv[1])
koppen_table = pd.read_csv(os.path.join('assets','koppen_table.csv'))
#==============================================================================
# Compute areas covered by major KG classes and transitions
#==============================================================================
scenarios = ['ssp119','ssp126','ssp245','ssp370','ssp434','ssp460','ssp585']
periods = config['periods_historical']+config['periods_future']
mapsize = config['upscale_mapsizes'][0]
# Area map (million km2)
res = 180/config['upscale_mapsizes'][0][0]
xi, yi = np.meshgrid(np.arange(-180+res/2,180+res/2,res), np.arange(90-res/2,-90-res/2,-res))
area_map = 10**-6*(40075*res/360)**2*np.cos(np.deg2rad(yi))
# Loop over scenarios and periods
df_kg_major_change_prct = pd.DataFrame(np.zeros((len(scenarios),2))*np.NaN,index=scenarios,columns=['1901-1930 to 1991-2020','1991-2020 to 2071-2100'])
for scenario in scenarios:
print('===============================================================================')
print('Compute areas covered by major KG classes and transitions for '+scenario)
kg_maps = np.zeros((mapsize[0],mapsize[1],len(periods)),dtype=np.single)*np.NaN
for pp in np.arange(len(periods)):
period = periods[pp]
# Load global Koppen-Geiger map
suffix = str(180/config['upscale_mapsizes'][0][0]).replace('.','p')
ncfile1 = os.path.join(config['folder_out'],'climatologies',str(period[0])+'_'+str(period[1]),'koppen_geiger_'+suffix+'.nc')
ncfile2 = os.path.join(config['folder_out'],'climatologies',str(period[0])+'_'+str(period[1]),scenario,'koppen_geiger_'+suffix+'.nc')
if os.path.isfile(ncfile1):
ncfile = ncfile1
elif os.path.isfile(ncfile2):
ncfile = ncfile2
else:
raise Exception("Unable to load map")
print('loading '+ncfile)
dset = Dataset(ncfile)
kg_class = np.array(dset.variables['kg_class'][:]).astype(int)
dset.close()
# Compute map of major classes
kg_major = np.zeros(kg_class.shape,dtype=int)
for ii in np.arange(koppen_table.shape[0]):
mask = kg_class==koppen_table['Class'][ii]
kg_major[mask] = koppen_table['Major'][ii]
# Insert map into array
kg_maps[:,:,pp] = kg_major
# Discard Antarctica
mask = np.min(kg_maps,axis=2)==0
for pp in np.arange(len(periods)):
kg_maps[:,:,pp][mask] = 0
# Make tables with area for major classes and transitions
df_kg_major_area_pct = pd.DataFrame(np.zeros((len(periods),6))*np.NaN,index=None,columns=['Period','A','B','C','D','E'])
df_kg_major_area_mm2 = pd.DataFrame(np.zeros((len(periods),6))*np.NaN,index=None,columns=['Period','A','B','C','D','E'])
df_transitions_mm2 = pd.DataFrame(np.zeros((1000,5))*np.NaN,index=None,columns=['From','To','Source','Target','Area'])
count = 0
for ll in np.arange(len(periods)):
# Compute areas
mask_land = kg_maps[:,:,0]!=0
df_kg_major_area_pct.iloc[ll,0] = str(periods[ll])
df_kg_major_area_mm2.iloc[ll,0] = str(periods[ll])
for cl in np.arange(1,6):
mask_cl = kg_maps[:,:,ll]==cl
df_kg_major_area_pct.iloc[ll,cl] = 100*np.sum(area_map[mask_cl])/np.sum(area_map[mask_land])
df_kg_major_area_mm2.iloc[ll,cl] = np.sum(area_map[mask_cl])
# Don't compute transitions for last period
if ll==len(periods)-1:
continue
# Loop over transitions
for source in np.arange(1,6):
for target in np.arange(1,6):
mask = ((kg_maps[:,:,ll]==source)==True) & ((kg_maps[:,:,ll+1]==target)==True)
mask_area = np.sum(area_map[mask])
# Add transition to table
if mask_area>0.05:
df_transitions_mm2.iloc[count,:] = np.array([str(periods[ll]),str(periods[ll+1]),source,target,mask_area])
count +=1
# Save results
df_kg_major_area_pct.to_csv(os.path.join(config['folder_stats'],'climatologies',scenario+'_kg_major_area_pct.csv'),index=False)
df_kg_major_area_mm2.to_csv(os.path.join(config['folder_stats'],'climatologies',scenario+'_kg_major_area_mm2.csv'),index=False)
df_transitions_mm2.to_csv(os.path.join(config['folder_stats'],'climatologies',scenario+'_transitions_mm2.csv'),index=False)
# Compute percentage of land surface that changes
mask_land = kg_maps[:,:,0]!=0
ll_1901 = np.array(periods)[:,0]==1901
ll_1991 = np.array(periods)[:,0]==1991
ll_2071 = np.array(periods)[:,0]==2071
diff1 = np.squeeze(np.abs(kg_maps[:,:,ll_1991]-kg_maps[:,:,ll_1901]))>0
diff2 = np.squeeze(np.abs(kg_maps[:,:,ll_2071]-kg_maps[:,:,ll_1991]))>0
diff1 = 100*np.mean(diff1[mask_land]*area_map[mask_land])/np.mean(area_map[mask_land])
diff2 = 100*np.mean(diff2[mask_land]*area_map[mask_land])/np.mean(area_map[mask_land])
df_kg_major_change_prct.loc[scenario] = [diff1,diff2]
# Save results
df_kg_major_change_prct.to_csv(os.path.join(config['folder_stats'],'climatologies','kg_major_change_prct.csv'),index=False)
#==============================================================================
# Load station data and compute KG classes
#==============================================================================
if os.path.isfile(os.path.join(config['folder_out'],'climatologies_validation','station_data.pickle')):
print('===============================================================================')
print('Loading existing station data file')
t = time.time()
station_data = pickle.load(open(os.path.join(config['folder_out'],'climatologies_validation','station_data.pickle'),'rb'))
print("Time elapsed is "+str(time.time()-t)+" sec")
if os.path.isfile(os.path.join(config['folder_out'],'climatologies_validation','station_data.pickle'))==False:
print('===============================================================================')
print('Loading station data')
dates_daily = pd.date_range(start=pd.to_datetime(datetime(1900,1,1)),end=pd.to_datetime('today'), freq='D')
station_files = glob.glob(os.path.join(config['folder_station'],'*.mat'))
station_data = {}
station_data['lat'] = np.zeros((len(station_files)),dtype=np.single)*np.NaN
station_data['lon'] = np.zeros((len(station_files)),dtype=np.single)*np.NaN
station_data['name'] = np.zeros((len(station_files)),dtype=object)*np.NaN
station_data['T_monthly_clim'] = np.zeros((len(station_files),len(config['periods_historical']),12),dtype=np.single)*np.NaN
station_data['P_monthly_clim'] = np.zeros((len(station_files),len(config['periods_historical']),12),dtype=np.single)*np.NaN
station_data['Class'] = np.zeros((len(station_files),len(config['periods_historical'])),dtype=np.single)*np.NaN
station_data['Major'] = np.zeros((len(station_files),len(config['periods_historical'])),dtype=np.single)*np.NaN
# Loop over stations
for ii in np.arange(len(station_files)):
print('Loading station '+str(ii))
t0 = time.time()
if ii in np.linspace(0,len(station_files),20).astype(int):
print(str(np.round(100*ii/len(station_files)))+' % completed')
# Read latitude, longitude, and name from mat file
station_data['lat'][ii] = tools.readmatfile(station_files[ii],'StationCoords/Lat')[0]
station_data['lon'][ii] = tools.readmatfile(station_files[ii],'StationCoords/Lon')[0]
station_data['name'][ii] = os.path.basename(station_files[ii])[:-4]
if np.isnan(station_data['lat'][ii]+station_data['lon'][ii]) | (abs(station_data['lat'][ii])>89.5) | (abs(station_data['lon'][ii])>179.5):
continue
# Read data from mat file
vars = ['PRCP','TMIN','TMAX','TAVG']
statdata = {}
for var in vars:
statdata[var] = np.zeros((len(dates_daily),1))*np.NaN
try:
statdata[var] = tools.readmatfile(station_files[ii],var).flatten().reshape(-1,1)
if len(statdata[var])<len(dates_daily):
statdata[var] = np.concatenate((statdata[var],np.zeros((len(dates_daily),1))*np.NaN),axis=0)
statdata[var] = statdata[var][:len(dates_daily)]
if (var=='PRCP'):
statdata[var] = statdata[var]*30.4 # Compute monthly total
if (var=='PRCP') & ("GSOD" in station_name[ii]):
statdata[var] = tools.eliminate_trailing_zeros(statdata[var]) # Eliminate erroneous zero precipitation in GSOD stations
except:
continue
# If TAVG not available, compute mean daily air temperature from TMIN and TMAX
sel = np.isnan(statdata['TAVG'])
statdata['TAVG'][sel] = ((statdata['TMIN']+statdata['TMAX'])/2)[sel]
# Compute monthly precipitation and air temperature climatologies and Koppen-Geiger class
for pp in np.arange(len(config['periods_historical'])):
period = config['periods_historical'][pp]
sel = (dates_daily>=datetime(period[0],1,1)) & (dates_daily<=datetime(period[1],12,31))
T_monthly_clim = tools.compute_monthly_climatology(statdata['TAVG'][sel],dates_daily[sel])
P_monthly_clim = tools.compute_monthly_climatology(statdata['PRCP'][sel],dates_daily[sel])
station_data['T_monthly_clim'][ii,pp,:] = T_monthly_clim
station_data['P_monthly_clim'][ii,pp,:] = P_monthly_clim
KG_dict = tools.koppen_geiger(T_monthly_clim.reshape(12,1,1),P_monthly_clim.reshape(12,1,1),koppen_table)
station_data['Class'][ii,pp] = KG_dict['Class']
station_data['Major'][ii,pp] = KG_dict['Major']
print(station_data['Major'][ii,:])
print("Time elapsed is "+str(time.time()-t0)+" sec")
if os.path.isdir(os.path.join(config['folder_out'],'climatologies_validation'))==False:
os.makedirs(os.path.join(config['folder_out'],'climatologies_validation'))
with open(os.path.join(config['folder_out'],'climatologies_validation','station_data.pickle'), 'wb') as f:
pickle.dump(station_data, f)
#==============================================================================
# Count number of stations for each provider
#==============================================================================
# Get list of providers
provider = []
for station_name in station_data['name']:
provider.append(station_name.split('_')[0])
providers = np.unique(provider)
print(providers)
# Get number of stations for each provider
for provider in providers:
res = [i for i in station_data['name'] if provider in i]
print(provider+' '+str(len(res))+' stations')
#==============================================================================
# Compute classification accuracy for each period for both 30 classes and
# for the major classes
#==============================================================================
print('===============================================================================')
print('Computing accuracy for historical periods')
nperiods = len(config['periods_historical'])
df_accuracy = pd.DataFrame(np.zeros((nperiods,6))*np.NaN,index=None,columns=['Period','nobs','Class','Major','conf_correct','conf_incorrect'])
for pp in np.arange(nperiods):
period = config['periods_historical'][pp]
print(period)
# Load global Koppen-Geiger map
suffix = str(180/config['mapsize'][0]).replace('.','p')
ncfile = os.path.join(config['folder_out'],'climatologies',str(period[0])+'_'+str(period[1]),'koppen_geiger_'+suffix+'.nc')
dset = Dataset(ncfile)
kg_class = np.array(dset.variables['kg_class'][:]).astype(int)
kg_confidence = np.array(dset.variables['kg_confidence'][:]).astype(int)
dset.close()
# Compute map of major classes
kg_major = np.zeros(kg_class.shape,dtype=int)
for ii in np.arange(koppen_table.shape[0]):
mask = kg_class==koppen_table['Class'][ii]
kg_major[mask] = koppen_table['Major'][ii]
# Convert lat/lon to row/col
ys = config['mapsize'][0]*(90-station_data['lat'])/180-0.5
ys = np.round(ys).astype(int)
ys[(ys<0) | (ys>=config['mapsize'][0])] = 0
xs = config['mapsize'][1]*(180+station_data['lon'])/360-0.5
xs = np.round(xs).astype(int)
xs[(xs<0) | (xs>=config['mapsize'][1])] = 0
# Compute accuracy using station data
valid = (np.isnan(station_data['Class'][:,pp])==False) \
& (kg_class[ys,xs]!=0) \
& (np.abs(station_data['lat'])<89.5) \
& (np.abs(station_data['lon'])<179.5) \
& (np.isnan(station_data['lat'])==False) \
& (np.isnan(station_data['lon'])==False)
df_accuracy['Period'][pp] = str(period)
df_accuracy['nobs'][pp] = np.sum(valid)
df_accuracy['Class'][pp] = 100*np.sum(station_data['Class'][:,pp][valid]==kg_class[ys,xs][valid])/np.sum(valid)
df_accuracy['Major'][pp] = 100*np.sum(station_data['Major'][:,pp][valid]==kg_major[ys,xs][valid])/np.sum(valid)
# Compute confidence level of correct and incorrect classifications
correct = (valid) & (station_data['Class'][:,pp]==kg_class[ys,xs])
incorrect = (valid) & (station_data['Class'][:,pp]!=kg_class[ys,xs])
df_accuracy['conf_correct'][pp] = np.mean(kg_confidence[ys,xs][correct])
df_accuracy['conf_incorrect'][pp] = np.mean(kg_confidence[ys,xs][incorrect])
# Performance comparison inside and outside US
US = (station_data['lat']<50) & (station_data['lat']>30) & (station_data['lon']<-66) & (station_data['lon']>-126)
sel1 = (valid) & (US)
sel2 = (valid) & (~US)
acc_US = 100*np.sum(station_data['Class'][:,pp][sel1]==kg_class[ys,xs][sel1])/np.sum(sel1)
acc_nonUS = 100*np.sum(station_data['Class'][:,pp][sel2]==kg_class[ys,xs][sel2])/np.sum(sel2)
print('Accuracy in US: '+str(acc_US)+' (n='+str(sum(sel1))+') Outside US: '+str(acc_nonUS)+' ('+str(sum(sel2))+')')
sel3 = (valid) & (np.isnan(np.mean(station_data['Class'],axis=1))==False)
acc_same = 100*np.sum(station_data['Class'][:,pp][sel3]==kg_class[ys,xs][sel3])/np.sum(sel3)
print('Accuracy same: '+str(acc_same))
# Clear memory
del kg_class
del kg_major
del kg_confidence
# Save to csv
print(df_accuracy)
df_accuracy.set_index('Period',inplace=True)
if os.path.isdir(os.path.join(config['folder_stats'],'validation'))==False:
os.makedirs(os.path.join(config['folder_stats'],'validation'))
df_accuracy.to_csv(os.path.join(config['folder_stats'],'validation','accuracy.csv'))
pdb.set_trace()
if __name__ == '__main__':
main()