-
Notifications
You must be signed in to change notification settings - Fork 0
/
pproc_ies.py
703 lines (567 loc) · 25.9 KB
/
pproc_ies.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
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
import os, sys, glob
import shutil
import pandas as pd
import numpy as np
import pyemu
from pymarthe import MartheModel
from pymarthe.utils import marthe_utils, shp_utils, pest_utils, pp_utils
from pymarthe.mfield import MartheField, MartheFieldSeries
from pymarthe.helpers.postprocessing import PestPostProcessing
import matplotlib.pyplot as plt
import matplotlib
from matplotlib.lines import Line2D
# plot settings
plt.rc('font', family='serif', size=9)
sgcol_width = 9/2.54
mdcol_width = 14/2.54
dbcol_width = 19/2.54
# ---------------------------------------------
# load pst and observation data from .config
# ---------------------------------------------
pstfile = 'cal_lizonne.pst'
pst = pyemu.Pst(pstfile)
case = pst.filename.split('.')[0]
hdic, pdics, odics = pest_utils.read_config(pst.filename.replace('.pst','.config'))
ogdates = { odic['locnme']:pd.to_datetime(odic['dates_out'].split('|')) for odic in odics}
# iteration id for posterior distribution
pt_id = pst.control_data.noptmax
pt_id = 3
# ---------------------------------------------
# load prior and last iteration observation ensembles
# ---------------------------------------------
# prior ensemble of simulated values
pr_oe = pyemu.ObservationEnsemble.from_csv(pst=pst,filename=os.path.join(f"{case}.0.obs.csv"))
pr_oe._df.index=pr_oe.index.astype(str) #trick since index mixed types in oe.index (should find a fix)
# posterior (last iteration) ensemble of simulated values
pt_oe = pyemu.ObservationEnsemble.from_csv(pst=pst,filename=os.path.join(f"{case}.{pt_id}.obs.csv"))
pt_oe._df.index=pt_oe.index.astype(str) #trick since index mixed types in oe.index (should find a fix)
# observation with noise
obswns = pyemu.ObservationEnsemble.from_csv(pst=pst,filename=os.path.join(f"{case}.obs+noise.csv"))
# ---------------------------------------------
# load prior and last iteration parameter ensembles
# ---------------------------------------------
# load parameter ensembles
pe_dic = { it:pyemu.ParameterEnsemble.from_csv(
pst=pst,filename=os.path.join(f"{case}.{it}.par.csv")
) for it in range(pt_id+1)}
# load prior parameter ensemble
pr_pe = pe_dic[0]
# load posterior parameter ensemble
pt_pe = pe_dic[pt_id]
# ---------------------------------------------
# save pe-oe for restart
# ---------------------------------------------
'''
nreals_restart = 116
best_reals = pr_oe.phi_vector.sort_values().index[0:nreals_restart]
pr_oe_best = pr_oe.loc[best_reals].copy()
pr_oe_best.to_csv(filename=f'{case}.restart.obs.csv')
pr_pe_best = pr_pe.loc[best_reals.astype(str)].copy()
pr_pe_best.to_csv(filename=f'{case}.restart.par.csv')
obswns_best = obswns.loc[best_reals].copy()
obswns_best.to_csv(filename=f'{case}.restart.obs+noise.csv')
pst.pestpp_options['ies_restart_parameter_ensemble'] = f'{case}.restart.par.csv'
pst.pestpp_options['ies_observation_ensemble'] = f'{case}.restart.obs+noise.csv'
pst.pestpp_options['ies_restart_observation_ensemble'] = f'{case}.restart.obs.csv'
pst.pestpp_options['ies_bad_phi'] = 1e10
pst.control_data.noptmax=2
pst.write(f'{case}.pst',version=2)
'''
# ---------------------------------------------
# filtering with total phi values
# ---------------------------------------------
# filter out bad phi
def filter_ens(pe,oe,qt,th):
pv = oe.phi_vector
th = min(pv.quantile(qt),th)
keep_reals = pv.loc[pv<th].index
return(pe.loc[keep_reals,:], oe.loc[keep_reals,:])
# remove extreme values from prior ensembles
qt=1.0
th=1e8
fpr_pe, fpr_oe = filter_ens(pr_pe,pr_oe,qt=qt,th=th)
'''
# extract and save filtered posterior ensembles
qt=0.5
th=1e8
fpt_pe, fpt_oe =filter_ens(pt_pe,pt_oe,qt=qt,th=th)
fpt_pe.to_csv(os.path.join(f"{case}.fpt.par.csv"))
'''
# ---------------------------------------------
# more granular filtering by phi contributions
# ---------------------------------------------
def get_phi_df(oe):
cols = oe._df.columns
pst = oe.pst
ogroups = pst.observation_data.loc[cols].groupby("obgnme").groups
res = pd.DataFrame(data={'name':cols,
'group':pst.observation_data.loc[cols,'obgnme'].values,
'modelled':np.nan,
'residual':np.nan
})
res.index = res.name
obs = pst.observation_data.loc[cols, ['obsval','weight']]
phi_rows = []
for idx in oe._df.index.values:
res.loc[cols,'modelled'] = oe._df.loc[idx, cols]
res.loc[cols,'residual'] = res.loc[cols,'modelled'] - obs.loc[cols,'obsval']
contribs = pyemu.Pst.get_phi_components(ogroups,res,obs,None,None)
phi_rows.append(contribs)
return(pd.DataFrame(data=phi_rows, index=oe.index, columns=ogroups))
phi_df = get_phi_df(pt_oe)
phi_df = phi_df.loc[:,~phi_df.columns.str.endswith('mf')]
# take best phi values by quantiles
phi_accept_by_ogroup = phi_df < phi_df.quantile(0.75)
# observation groups of absolute head values
hlocs = phi_df.columns[phi_df.columns.str.contains('x') & ~phi_df.columns.str.endswith('mf')]
# filter
phi_accept = phi_accept_by_ogroup[hlocs].apply(lambda x : np.logical_and.reduce(x), axis=1)
keep_reals = phi_df.index[phi_accept]
# append base realization
keep_reals = pd.Index.union(keep_reals,["base"])
fpt_pe = pt_pe.loc[keep_reals,:]
fpt_oe = pt_oe.loc[keep_reals,:]
fpt_pe.to_csv('cal_lizonne.fpt.par.csv')
# ---------------------------------------------
# evolution of phi and hist for last iteration
# ---------------------------------------------
fig, axes = plt.subplots(1, 2, sharey=True, figsize=(10,3.5))
# left
ax = axes[0]
phi = pd.read_csv(os.path.join(f"{case}.phi.actual.csv"),index_col=0)
phi.index = phi.total_runs
phi.iloc[:,6:].apply(np.log10).plot(legend=False,lw=0.5,color='k', ax=ax)
ax.set_title(r'Actual ')
ax.set_ylabel(r'log ')
# right
ax = axes[-1]
phi = pd.read_csv(os.path.join(f"{case}.phi.meas.csv"),index_col=0)
phi.index = phi.total_runs
phi.iloc[:,6:].apply(np.log10).plot(legend=False,lw=0.2,color='r', ax=ax)
ax.set_title(r'Measured+Noise ')
fig.tight_layout()
fig.savefig(os.path.join('pproc','phi_evol.pdf'),dpi=300)
# ---------------------------------------------
# compare phi distrib for prior and last iter
# ---------------------------------------------
pr_logphi = pr_oe.phi_vector.apply(np.log10)
pt_logphi = pt_oe.phi_vector.apply(np.log10)
bins=np.histogram(np.hstack((pr_logphi)), bins=40)[1] #get the bin edges
bins=np.histogram(np.hstack((pr_logphi,pt_logphi)), bins=40)[1] #get the bin edges
fig,ax = plt.subplots(1,1,figsize=(3.5,3.5))
pr_logphi.hist(bins=bins,ax=ax,fc="0.5",ec="none",alpha=0.5,density=False,label='prior')
pt_logphi.hist(bins=bins,ax=ax,fc="b",ec="none",alpha=0.5,density=False,label='posterior')
ax.legend()
_ = ax.set_xlabel('log10($\Phi$)')
fig.tight_layout()
fig.savefig(os.path.join('pproc','phi.pdf'),dpi=300)
# ---------------------------------------------
# time series
# ---------------------------------------------
# get ensemble of time series for given observation group
def get_og_ts(oe,onames,odates, trans):
ts = oe._df.loc[:,onames].T.apply(trans)
ts.index = odates
return(ts)
def plot_tseries_ensembles(pr_oe, pt_oe, obswns, ognmes, ogdates, trans=None, ylabel='',legend=True ):
# get the observation data from the control file and select
obs = pst.observation_data.copy()
fig,axes = plt.subplots(len(ognmes),1,sharex=True,figsize=(dbcol_width,0.8*dbcol_width))
if trans==None:
trans=[lambda x : x]*len(ognmes)
if not type(trans)==list:
trans = [trans]*len(ognmes)
# for each observation group (i.e. timeseries)
for ax,og,t in zip(axes,ognmes,trans):
# get values
oobs = obs.loc[obs.obgnme==og.lower(),:].copy()
onames = oobs.obsnme.values
odates = ogdates[og]
# plot prior
if pr_oe is not None :
ts = get_og_ts(pr_oe,onames,odates, trans=t)
ts.plot(ax=ax,color='grey',lw=0.5,alpha=0.10,legend=False)
# plot posterior
if pt_oe is not None :
ts = get_og_ts(pt_oe,onames,odates,trans=t)
ts.plot(ax=ax,color='red',lw=0.5,alpha=0.50,legend=False)
ts['base'].plot(ax=ax,color='green',alpha=1,lw=2,legend=False)
# plot measured+noise
if obswns is not None :
ts = get_og_ts(obswns,onames,odates,trans=t)
ts.plot(ax=ax,color='blue',lw=0.5,alpha=0.05,legend=False)
# plot obs
ax.plot(odates, oobs.obsval.apply(t).values,color="black",alpha=1,lw=1)
ax.set_title(og,loc="left")
ax.set_ylabel(ylabel)
lpr = Line2D([0], [0], label='Sim. prior', color='grey')
lpt = Line2D([0], [0], label='Sim. posterior', color='red')
lbase = Line2D([0], [0], label='Sim. base', color='green')
lobs = Line2D([0], [0], label='Observed', color='black')
lobsn = Line2D([0], [0], label='Obs.+noise', color='blue')
if legend:
ax.legend(handles=[lpr,lpt,lbase,lobs,lobsn],loc='upper left',ncols=5)
plot_legend=False
fig.tight_layout()
return fig
# gaging stations
gstations = ['P7250001','P7270001','P8215010','P8284010']
fig = plot_tseries_ensembles(pr_oe, pt_oe, obswns , gstations, ogdates,trans=lambda x : 10**x, ylabel='River discharge [m$^3$/s]')
fig.savefig(os.path.join('pproc','pr_pt_qsimobs.png'),dpi=300)
fig = plot_tseries_ensembles(None, pt_oe, obswns , gstations, ogdates,trans=lambda x : 10**x, ylabel='River discharge [m$^3$/s]')
fig.savefig(os.path.join('pproc','pt_qsimobs.png'),dpi=300)
fig = plot_tseries_ensembles(None, fpt_oe, obswns , gstations, ogdates,trans=lambda x : 10**x, ylabel='River discharge [m$^3$/s]')
fig.savefig(os.path.join('pproc','fpt_qsimobs.png'),dpi=300)
# selection of observation wells
obswells = ['07333X0027','07345X0023','07346X0017','07346X0083','07574X0014']
fig = plot_tseries_ensembles(pr_oe, pt_oe, obswns , obswells, ogdates,trans=lambda x : x, ylabel='Water level [m NGF]')
fig.savefig(os.path.join('pproc','pr_pt_hsimobs.png'),dpi=300)
fig = plot_tseries_ensembles(None, pt_oe, obswns , obswells, ogdates,trans=lambda x : x, ylabel='Water level [m NGF]')
fig.savefig(os.path.join('pproc','pt_hsimobs.png'),dpi=300)
fig = plot_tseries_ensembles(None, fpt_oe, obswns , obswells, ogdates,trans=lambda x : x, ylabel='Water level [m NGF]')
fig.savefig(os.path.join('pproc','fpt_hsimobs.png'),dpi=300)
# --------- selection of the two (article)
obslocs = pd.read_excel(os.path.join('..','data','SIG','obsloc_labels.xlsx'),index_col='id')
locs = ['P8284010','P7250001','07333X0027','07346X0017']
labels = obslocs.loc[locs].label
units = ['m$^3$/s']*2 + ['m a.s.l.']*2
trans = [lambda x : 10**x,lambda x : 10**x,lambda x : x,lambda x : x]
fig = plot_tseries_ensembles(None, fpt_oe, obswns , locs, ogdates,trans=trans, ylabel='',legend=False)
cal_start = pd.to_datetime('2012-08-01')
cal_end =pd.to_datetime('2019-07-31')
fig.axes[0].set_xlim(cal_start,cal_end)
fig.axes[0].set_xticks([])
fig.axes[0].set_xticks([],minor=True)
fig.axes[0].xaxis.set_major_locator(matplotlib.dates.MonthLocator(bymonth=1,bymonthday=1))
fig.axes[0].xaxis.set_major_formatter(matplotlib.dates.DateFormatter('%Y-%m'))
fig.axes[0].xaxis.set_minor_locator(matplotlib.dates.MonthLocator())
fig.axes[-1].tick_params(axis='x', which='major', pad=0)
for ax,unit,label in zip(fig.axes,units,labels):
ax.set_ylabel(unit)
ax.set_title(f' {label}',loc="left",y=0.80)
fig.axes[0].set_ylim(0,45)
fig.axes[1].set_ylim(0,5)
fig.axes[2].set_ylim(94,135)
#fig.axes[3].set_ylim(100,175)
lpt = Line2D([0], [0], label='Posterior', color='red')
lbase = Line2D([0], [0], label='Center', color='green')
lobs = Line2D([0], [0], label='Obs.', color='black')
lobsn = Line2D([0], [0], label='Obs.+noise', color='blue')
fig.axes[0].legend(handles=[lpt,lbase,lobs,lobsn],loc='upper right',ncols=4,fontsize=11)
fig.axes[0].margins(x=0)
fig.align_ylabels()
fig.tight_layout()
fig.subplots_adjust(hspace=0.10)
fig.savefig(os.path.join('pproc','fpt_q_and_h_simobs.png'),dpi=150)
fig.savefig(os.path.join('pproc','fpt_q_and_h_simobs.pdf'),dpi=150)
# ---------------------------------------------
# final ensemble selection
# ---------------------------------------------
# see Shafii et al 2015 : http://dx.doi.org/10.1016/j.jhydrol.2015.01.051
# compute reliability
def get_reliability(pt_oe, obs):
pt_oe = pt_oe
lqt = pt_oe._df.quantile(qt_lb)
uqt = pt_oe._df.quantile(qt_ub)
obs_in_bds = obs.between(lqt,uqt)
# re= (number of obs in pred. interval)/(total number of obs).
r = obs_in_bds.sum()/obs_in_bds.shape[0]
return(r)
# compute (normalized) sharpness
def get_sharpness(pt_oe,pr_oe=None):
pt_wdth = pt_oe._df.quantile(qt_ub)-pt_oe._df.quantile(qt_lb)
# When pr_oe is not provided, return absolute sharpness value
# (not relevant for heterogeneous observational dataset).
if pr_oe is None:
s = pt_wdth.mean()
# When pr_oe is provided, return normalized sharpness indicator
# s=1 for single value
# s=0 for prediction interval equal to posterior interval
# see
else :
pr_wdth = pr_oe._df.quantile(qt_ub)-pr_oe._df.quantile(qt_lb)
wdth_ratio = 1 - pt_wdth/pr_wdth
s = wdth_ratio.mean()
return(s)
qt_lb=0.05
qt_ub=0.95
# -----------------------------------------------------
# --- reliability and sharpness analysis
# -----------------------------------------------------
# --- plot reliability and sharpness through iterations
fig,ax=plt.subplots(1,1,figsize=(6,4),sharey=True)
nit = phi.shape[0]
# all obs
rvals,svals,dvals=[],[],[]
for i in range(nit):
# load ensemble of simulated observations at IES iteration i
ioe = pyemu.ObservationEnsemble.from_csv(pst=pst,filename=os.path.join(f"{case}.{i}.obs.csv"))
# compute sharpness and reliability considering all subset
r=get_reliability(ioe,obswns.loc['base'])
s= get_sharpness(ioe,pr_oe)
d = np.sqrt((1-r)**2+(1-s)**2)
rvals.append(r)
svals.append(s)
dvals.append(d)
rsd_iter = rvals, svals, dvals
ax.plot(rvals,'--+', color='royalblue',label='Reliability')
ax.plot(svals,'--+', color='darkgreen',label='Sharpness')
ax.plot(dvals,'-+', color='black', label='Distance to optimum')
ax.axvline(3,color='grey')
ax.set_xlabel('IES iterations')
ax.legend('lower right')
fig.savefig(os.path.join('pproc','rsd_through_iterations.pdf'),dpi=300)
# --- plot reliability and sharpness through final filtered ensemble size
fig,ax=plt.subplots(1,1,figsize=(6,4),sharey=True)
# posterior observation ensemble sorted by decreasing phi
spt_oe = fpt_oe.loc[fpt_oe.phi_vector.sort_values(ascending=False).index]
# compute distance to optimum (r=1,s=1) to get optimum ffpt ensemble size
# all obs.
rvals,svals,dvals=[],[],[]
for n in range(spt_oe.shape[0]):
r=get_reliability(spt_oe.iloc[:n+1],obswns.loc['base'])
s= get_sharpness(spt_oe.iloc[:n+1],pr_oe)
d = np.sqrt((1-r)**2+(1-s)**2)
rvals.append(r)
svals.append(s)
dvals.append(d)
ax.plot(rvals,'--+', color='royalblue',label='Reliability')
ax.plot(svals,'--+', color='darkgreen',label='Sharpness')
ax.plot(dvals,'-+', color='black', label='Distance to optimum')
ax.set_xlabel('Number of realizations in the final filtered posterior ensemble')
ax.legend('lower right')
fig.tight_layout()
fig.savefig(os.path.join('pproc','rsd_through_ensemble_size.pdf'),dpi=300)
# optimum size for the final filtered posterior ensemble
nreals_ffpt_ens=np.argmin(dvals)
print(f'Optimum number of best realizations is: {nreals_ffpt_ens}')
# final ensemble selection
ffpt_oe = spt_oe.iloc[:nreals_ffpt_ens] # set to nreals_ffpt_ens after examination of rsd.pdf
ffpt_ids = ffpt_oe.index
ffpt_pe = pt_pe.loc[ffpt_ids]
# identify new base (center) obs realization from normalized distance to mean, and update index
center_real = (ffpt_oe - ffpt_oe.mean()).div(ffpt_oe.mean()).apply(np.linalg.norm, axis=1).idxmin()
ffpt_pe._df.index = ffpt_pe.index.str.replace(center_real,'ffpt_center')
ffpt_oe._df.index = ffpt_oe.index.str.replace(center_real,'ffpt_center')
# write final parameter ensemble
ffpt_pe.to_csv('cal_lizonne.ffpt.par.csv')
# plot final filtered ensemble
fig = plot_tseries_ensembles(None, ffpt_oe, obswns , gstations, ogdates,trans=lambda x : 10**x, ylabel='River discharge [m$^3$/s]')
fig.savefig(os.path.join('pproc','ffpt_qsimobs.png'),dpi=300)
fig = plot_tseries_ensembles(None, ffpt_oe, obswns , obswells, ogdates,trans=lambda x : x, ylabel='Water level [m NGF]')
fig.savefig(os.path.join('pproc','ffpt_hsimobs.png'),dpi=300)
# ---------------------------------------------
# evolution of phi and hist for last iteration + rsd
# ---------------------------------------------
fig, axes = plt.subplots(1, 2, figsize=(dbcol_width,0.4*dbcol_width),width_ratios=[6.5,4])
# phi evol
phi_color = 'grey'
ax = axes[0]
phi = pd.read_csv(os.path.join(f"{case}.phi.actual.csv"),index_col=0)
phi.iloc[:,6:].apply(np.log10).plot(legend=False,lw=0.5,color=phi_color,alpha=0.8, ax=ax)
ax.set_ylabel('log10($\Phi$)')
ax.set_xlabel('IES iterations')
ax.axvline(0,color='grey',ls='--',lw=3,alpha=0.6)
ax.axvline(3,color='darkred',ls='--',lw=3,alpha=0.6)
ax.text(0.5,7.4,'a)',fontsize=20)
phi_evol_handle = Line2D([0], [0], label='$\Phi$ (232 realizations)', color=phi_color, marker=None, linestyle= '-')
ax.legend(handles=[phi_evol_handle], alignment='left', loc='lower left',fontsize=9, bbox_to_anchor=(0.08, 0))
# rsd
twax= ax.twinx()
rsd_color = 'blue'
rvals, svals, dvals = rsd_iter
rl = twax.plot(rvals,'--+', color='royalblue',label='Reliability')
sl = twax.plot(svals,'--*', color='darkgreen',label='Sharpness')
dl = twax.plot(dvals,'-', color='black', label='Distance/opt.')
twax.set_ylabel('Performance ratios [-]')
# legend
lns = rl + sl + dl
labs = [l.get_label() for l in lns]
twax.legend(lns, labs,loc='upper right', bbox_to_anchor=(1.01, 1.01),alignment='left',fontsize=9)
ax.set_ylabel('log10($\Phi$)')
# histogram
ax = axes[1]
bins=np.histogram(np.hstack((pr_logphi,pt_logphi)), bins=40)[1] #get the bin edges
pt_logphi.hist(bins=bins,ax=ax,fc="darkred",ec="none",alpha=0.6,density=False,label='Posterior (It. 3)')
pr_logphi.hist(bins=bins,ax=ax,fc="0.5",ec="none",alpha=0.6,density=False,label='Prior (It. 0)')
ax.yaxis.tick_right()
ax.yaxis.set_label_position("right")
ax.legend(loc='upper right')
ax.set_ylabel('Frequency')
_ = ax.set_xlabel('log10($\Phi$)')
ax.text(5.05,42,'b)',fontsize=20)
fig.subplots_adjust(wspace=0.30)
fig.tight_layout()
fig.savefig(os.path.join('pproc','phi_rsd.pdf'),dpi=300)
# ---------------------------------------------
# fit statistics
# ---------------------------------------------
# get all metrics
m = pyemu.utils.metrics.calc_metric_ensemble(ffpt_oe._df,pst)
# drop fluctuations
m = m.loc[:,~m.columns.str.endswith('mf')]
# multiindex
locs = [cname.split('_')[-1] for cname in m.columns]
metrics = ['_'.join(cname.split('_')[:-1]) for cname in m.columns]
m.columns = pd.MultiIndex.from_frame(
pd.DataFrame(
{'metric':metrics,
'loc':locs}
)
)
# label dics
label_dic = {id.lower():label for id,label in zip(obslocs.index,obslocs.label)}
m.columns = m.columns.set_levels(m.columns.levels[1].map(label_dic,na_action='ignore'),level=1)
allgstations = obslocs.loc[obslocs.index.str.startswith('P'),'label'].sort_values()
allobswells = obslocs.loc[obslocs.index.str.contains('X'),'label'].sort_values()
fig, axs = plt.subplots(1,2,figsize=(dbcol_width,0.33*dbcol_width),width_ratios=[len(allgstations),len(allobswells)])
ax1 = m.loc[:,('KGE',allgstations)].boxplot(showfliers=False,ax=axs[0])
ax1.set_xticklabels(allgstations.values)
ax1.set_ylabel('KGE [-]')
ax1.set_title('Gaging stations (KGE)')
ax2 = m.loc[:,('RMSE',allobswells)].boxplot(showfliers=False, ax=axs[1])
ax2.set_xticklabels(allobswells.values)
ax2.set_title('Observation wells (RMSE)')
ax2.set_ylabel('RMSE [m]')
fig.tight_layout()
fig.savefig(os.path.join('pproc','metrics_boxplots.pdf'),dpi=300)
# ---------------------------------------------
# parameters evolution
# ---------------------------------------------
dic_props = dict(color = 'k', lw = 0.5)
meanprops = dict(marker = 'o', mfc = 'none', ms = 3, lw = 0.5, mec = 'k')
flierprops = dict(marker = 'o', mfc = 'none', ms = 3, lw = 0.1, mec = 'k')
def plot_multviolins(ens_list, ax, showpoints=False):
bplot = ax.boxplot(ens_list, widths = 0.4, showbox = True, showcaps = False,
showmeans = True, showfliers = False, boxprops = dic_props,
whiskerprops = dic_props, capprops = dic_props, medianprops = dic_props,
meanprops = meanprops, flierprops = flierprops)
# violin plot (white)
vp = ax.violinplot(ens_list, widths = 0.8, points = 100, showmeans = False,
showmedians = False, showextrema = False)
for pc in vp['bodies']:
pc.set_facecolor('white')
pc.set_edgecolor('white')
pc.set_alpha(1)
# violin plot (color)
vp = ax.violinplot(ens_list, widths = 0.8, points = 100, showmeans = False,
showmedians = False, showextrema = False)
for i, pc in enumerate(vp['bodies']):
pc.set_facecolor('blue')
pc.set_edgecolor('blue')
pc.set_alpha(0.4)
if showpoints != False: # Points
for i, tick in enumerate(xticks_list):
y = ens_list[i]
x = np.random.normal(tick, 0.04, size = len(y))
ax.plot(x, y, 'k.', alpha = 0.1)
ax.set_xticklabels([])
ax.set_axisbelow(True)
ax.yaxis.grid()
par = pst.parameter_data
df_list = []
for i in range(pt_id+1):
pe_df = pe_dic[i]._df.copy(deep=True)
pe_df.columns = pd.MultiIndex.from_frame(
pd.DataFrame({
'pargp':par.loc[pe_df.columns,'pargp'].values,
'layer':pe_df.columns.str.extract(r"_l(\d+)_z",expand=False).astype(float)-1,
'parnme':pe_df.columns.values
}))
df_list.append(pe_df)
pes = pd.concat(df_list,keys=range(pt_id+1))
pfs =['emmca','emmli','permh']
for prefix in pfs:
fig, axs = plt.subplots(6,1,figsize=(10,18))
for l in range(6):
parvals_list=[pes.loc[(i,slice(None)),\
(pes.columns.get_level_values(0).str.startswith(prefix),l,slice(None))
].values.ravel() \
for i in range(pt_id+1)]
ax = plot_multviolins(parvals_list, axs[l])
axs[l].set_title(f'{prefix} for layer {l+1}')
fig.tight_layout()
fig.savefig(os.path.join('pproc',f'evol_{prefix}.pdf'),dpi=300)
plt.close(fig)
pfs =['cap_sol_progr', 'equ_ruis_perc', 't_demi_percol','perm_r_zpc']
fig, axs = plt.subplots(len(pfs),1,figsize=(10,18))
for i,prefix in enumerate(pfs):
parvals_list=[pes.loc[(i,slice(None)),\
(pes.columns.get_level_values(0).str.startswith(prefix),slice(None),slice(None))
].values.ravel() \
for i in range(pt_id+1)]
ax = plot_multviolins(parvals_list, axs[i])
axs[i].set_title(f'{prefix}')
fig.tight_layout()
fig.savefig(os.path.join('pproc',f'evol_other_params.pdf'),dpi=300)
plt.close(fig)
# ---------------------------------------------
# parameters pt - pr distributions
# ---------------------------------------------
# get parameter df and append phi value value
pr_pe_df = pr_pe._df.copy(deep=True)
pt_pe_df = pt_pe._df.copy(deep=True)
# set multiindexed columns with parmameter groups (for hist grouping)
par = pst.parameter_data
parnmes= pr_pe_df.columns
colmix = pd.MultiIndex.from_frame(
pd.DataFrame({
'pargp':par.loc[parnmes,'pargp'],
'parnme':parnmes
}))
pr_pe_df.columns = colmix
pt_pe_df.columns = colmix
# plot hydraulic properties (per layer)
pfs =['emmca','emmli','permh','perm_r']
for prefix in pfs:
pr_df = pr_pe_df.loc[:,(pr_pe_df.columns.get_level_values(0).str.startswith(prefix),slice(None))].melt()
pt_df = pt_pe_df.loc[:,(pt_pe_df.columns.get_level_values(0).str.startswith(prefix),slice(None))].melt()
axs = pr_df.hist(by='pargp',fc="0.5",ec="none",alpha=0.5,density=False,label='prior')
axs = pt_df.hist(ax=axs,by='pargp',fc="b",ec="none",alpha=0.5,density=False,label='posterior')
try :
fig = axs[0].get_figure()
except :
fig = axs.get_figure()
fig.tight_layout()
fig.savefig(os.path.join('pproc',f'pr_pt_hist_{prefix}.pdf'),dpi=300)
# plot soil parameters
pfs =['cap_sol_progr', 'equ_ruis_perc', 't_demi_percol']
fig,axs=plt.subplots(1,3,figsize=(12,6))
for i,prefix in enumerate(pfs):
pr_df = pr_pe_df.loc[:,(pr_pe_df.columns.get_level_values(0).str.startswith(prefix),slice(None))].melt()
pt_df = pt_pe_df.loc[:,(pt_pe_df.columns.get_level_values(0).str.startswith(prefix),slice(None))].melt()
ax = pr_df.hist(ax=axs[i],by='pargp',fc="0.5",ec="none",alpha=0.5,density=False,label='prior')
ax = pt_df.hist(ax=axs[i],by='pargp',fc="b",ec="none",alpha=0.5,density=False,label='posterior')
fig.tight_layout()
fig.savefig(os.path.join('pproc','pr_pt_hist_surf.pdf'),dpi=300)
# ---------------------------------------------
# extract realization
# ---------------------------------------------
# best real
real = pt_oe.index[pt_oe.phi_vector.argmin()]
# user defined real
par_set = pt_pe.loc[str(real)]
pst.parameter_data.loc[par_set.index,'parval1']=par_set.values
pst.control_data.noptmax=0
pst.write(f'caleval_r{real}_it{pt_id}_best.pst')
# write pst with posterior base pe
real = 'ffpt_center'
par_set = ffpt_pe.loc[real]
pst.parameter_data.loc[par_set.index,'parval1']=par_set.values
pst.control_data.noptmax=0
pst.write('caleval_ffpt_center.pst')
'''
# run pest with center real from final filtered posterior ensemble (ffpt)
pyemu.os_utils.run('pestpp-ies caleval_ffpt_center.pst')
# copy model files with center filtered posterior real to sim dir
tpl_sim_dir = os.path.join('..','tpl_sim')
# post-proc center real
print('post processing center realization...'
exec(open('../pproc_cal.py').read())
print('Copying model files of center real to {tpl_sim_dir}')
for f in glob.glob("Lizonne.*"): shutil.copy(f, tpl_sim_dir)
# copying the calibration heads (they will be used to define the constraints on gw level)
shutil.copy('chasim.out',os.path.join(tpl_sim_dir,'chasim.out.cal'))
# copying the final filtered posterior parameter ensemble (it will constitute the par stack for MOU)
shutil.copy('cal_lizonne.ffpt.par.csv',os.path.join(tpl_sim_dir,'cal_lizonne.ffpt.par.csv'))
'''