forked from erinwalker1/vert_res_attribution
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Fig7_return_period.py
242 lines (208 loc) · 10.2 KB
/
Fig7_return_period.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
# -*- coding: utf-8 -*-
"""
Created on Wed Mar 17 16:36:38 2021
Updated on Wed Jul 21 12:14:00 2021
@author: Erin Walker
"""
"""
Adapted from Sarah Sparrow Code:
https://github.com/snsparrow/Attribution_workshop.git
"""
import matplotlib.pyplot as plt
import numpy as np
import xarray as xr
import random
from scipy import stats
import pandas as pd
"""
P = Present
F = 3Deg
C = Control
DA = Double All
DT = Double Tropopause
Q = Quadruple All
NS = No Stratopshere
"""
filepath = "" # Filepath to southern UK Isca DJF precipitation
p_file_c = xr.open_dataset(filepath + "present_control_suk_pr_djf_all_ens.nc",chunks={'time':30})
f_file_c = xr.open_dataset(filepath + "future_control_suk_pr_djf_all_ens.nc")
p_file_da = xr.open_dataset(filepath + "present_double_all_suk_pr_djf_all_ens.nc",chunks={'time':30})
f_file_da = xr.open_dataset(filepath + "future_double_all_suk_pr_djf_all_ens.nc")
p_file_dt = xr.open_dataset(filepath + "present_double_tropo_suk_pr_djf_all_ens.nc",chunks={'time':30})
f_file_dt = xr.open_dataset(filepath + "future_double_tropo_suk_pr_djf_all_ens.nc")
p_file_q = xr.open_dataset(filepath + "present_quadruple_all_suk_pr_djf_all_ens.nc",chunks={'time':30})
f_file_q = xr.open_dataset(filepath + "future_quadruple_all_suk_pr_djf_all_ens.nc")
p_file_ns = xr.open_dataset(filepath + "present_no_strat_suk_pr_djf_all_ens.nc",chunks={'time':30})
f_file_ns = xr.open_dataset(filepath + "future_no_strat_suk_pr_djf_all_ens.nc")
# Load precipitation
p_c = p_file_c['precipitation'][:]
p_da = p_file_da['precipitation'][:]
p_dt = p_file_dt['precipitation'][:]
p_q = p_file_q['precipitation'][:]
p_ns =p_file_ns['precipitation'][:]
f_c = f_file_c['precipitation'][:]
f_da = f_file_da['precipitation'][:]
f_dt = f_file_dt['precipitation'][:]
f_q = f_file_q['precipitation'][:]
f_ns = f_file_ns['precipitation'][:]
#Resample for 2 daily sum
p_c_2day = p_c.resample(time='2D').sum(dim='time')
p_da_2day = p_da.resample(time='2D').sum(dim='time')
p_dt_2day = p_dt.resample(time='2D').sum(dim='time')
p_q_2day = p_q.resample(time='2D').sum(dim='time')
p_ns_2day = p_ns.resample(time='2D').sum(dim='time')
f_c_2day = f_c.resample(time='2D').sum(dim='time')
f_da_2day = f_da.resample(time='2D').sum(dim='time')
f_dt_2day = f_dt.resample(time='2D').sum(dim='time')
f_q_2day = f_q.resample(time='2D').sum(dim='time')
f_ns_2day = f_ns.resample(time='2D').sum(dim='time')
#Calculate two daily max for each ensemble
p_twoday_max_c = []
p_twoday_max_da = []
p_twoday_max_dt = []
p_twoday_max_q = []
p_twoday_max_ns = []
f_twoday_max_c = []
f_twoday_max_da = []
f_twoday_max_dt = []
f_twoday_max_q = []
f_twoday_max_ns = []
for ens in range(0,len(p_c.ensemble)):
p_twoday_max_c.append(p_c_2day.mean(('lat','lon'))[ens].max().values.item())
p_twoday_max_da.append(p_da_2day.mean(('lat','lon'))[ens].max().values.item())
p_twoday_max_dt.append(p_dt_2day.mean(('lat','lon'))[ens].max().values.item())
p_twoday_max_q.append(p_q_2day.mean(('lat','lon'))[ens].max().values.item())
p_twoday_max_ns.append(p_ns_2day.mean(('lat','lon'))[ens].max().values.item())
f_twoday_max_c.append(f_c_2day.mean(('lat','lon'))[ens].max().values.item())
f_twoday_max_da.append(f_da_2day.mean(('lat','lon'))[ens].max().values.item())
f_twoday_max_dt.append(f_dt_2day.mean(('lat','lon'))[ens].max().values.item())
f_twoday_max_q.append(f_q_2day.mean(('lat','lon'))[ens].max().values.item())
f_twoday_max_ns.append(f_ns_2day.mean(('lat','lon'))[ens].max().values.item())
#calculate percentiles
x2 = np.array(p_twoday_max_c).flatten()
x2.sort()
i2_95 = int(0.95 * len(x2))
i2_90 = int(0.90 * len(x2))
c2_95 = x2[i2_95]
c2_90 = x2[i2_90]
print(c2_90)
#Now calc return plots
def calc_return_times(em, direction="ascending", period=1):
ey_data = np.array(em).flatten()
ey_data.sort()
# reverse if necessary
if direction == "descending": # being beneath a threshold value
ey_data = ey_data[::-1]
# create the n_ens / rank_data
val = float(len(ey_data)) * 1.0/period
end = float(len(ey_data)) * 1.0/period
start = 1.0
step = (end - start) / (len(ey_data)-1)
ranks = [x*step+start for x in range(0, len(ey_data))]
ex_data = val / np.array(ranks, dtype=np.float32)
return ey_data, ex_data
def calc_return_time_confidences(em, direction="ascending", c=[0.05, 0.95], bsn=1e5):
# c = confidence intervals (percentiles) to calculate
# bsn = boot strap number, number of times to resample the distribution
ey_data = np.array(em).flatten()
# create the store
sample_store = np.zeros((int(bsn), ey_data.shape[0]), 'f')
# do the resampling
for s in range(0, int(bsn)):
t_data = np.zeros((ey_data.shape[0]), 'f')
for y in range(0, ey_data.shape[0]):
x = random.uniform(0, ey_data.shape[0])
t_data[y] = ey_data[int(x)]
t_data.sort()
# reverse if necessary
if direction == "descending":
t_data = t_data[::-1]
sample_store[s] = t_data
# now for each confidence interval find the value at the percentile
conf_inter = np.zeros((len(c), ey_data.shape[0]), 'f')
for c0 in range(0, len(c)):
for y in range(0, ey_data.shape[0]):
data_slice = sample_store[:,y]
conf_inter[c0,y] = stats.scoreatpercentile(data_slice, c[c0]*100)
return conf_inter
def find_nearest(array,value):
idx=(np.abs(array-value)).argmin()
return idx
def plot_return_time(exp,variable,dataAct,dataNat,threshold,dirn,fig,row,col):
# Setup the plot parameters and axis limits
ax = plt.subplot2grid((2,3),(row,col))
plt.title(exp,fontsize=16)
ax.set_ylabel(variable,fontsize=16)
ax.set_xlabel("Chance of event occurring in a given year",fontsize=16)
plt.setp(ax.get_xticklabels(),fontsize=16)
plt.setp(ax.get_yticklabels(),fontsize=16)
ax.set_xlim(1,1e2)
ax.set_ylim(16,45)
# Plot the return time for the historical and historicalNat simulations
plot_rt(dataAct,["royalblue","cornflowerblue","mediumblue"],"Present",ax,"both",dirn,threshold,"",'--')
plot_rt(dataNat,["orange","gold","darkorange"],"3Deg",ax,"both",dirn,threshold,"Threshold",'--')
labels=['','1/1','1/10','1/100']
ax.set_xticklabels(labels)
plt.setp(ax.get_xticklabels(),fontsize=16)
plt.setp(ax.get_yticklabels(),fontsize=16)
ll=ax.legend(loc="upper left",prop={"size": 14},fancybox=True,numpoints=1)
def plot_rt(data,cols,plabel,ax,errb,dirn,threshold,tlabel,tstyle):
# Plot the return times with bootstrap 9-95% confience intervals.
# Calculate the return times
y_data_all, x_data_all = calc_return_times(data,direction=dirn,period=1)
# Calculate the bootstrap confidences in both the x and y directions
conf_all = calc_return_time_confidences(data,direction=dirn,bsn=1000)
conf_all_x = calc_return_time_confidences(x_data_all,direction="descending",bsn=1000)
# Plot the return time curve
l1=ax.semilogx(x_data_all,y_data_all, marker='o',markersize=4,
linestyle='None',mec=cols[0],mfc=cols[0],
color=cols[0],fillstyle='full',
label=plabel,zorder=2)
conf_all_5=conf_all[0,:].squeeze()
conf_all_95=conf_all[1,:].squeeze()
conf_all_x_5=conf_all_x[0,:].squeeze()
conf_all_x_95=conf_all_x[1,:].squeeze()
#ax.grid(b=True,which='major')
#ax.grid(b=True, which='minor',linestyle='--')
# Plot the error bars onn the return times
#if errb=="both":
# cl0=ax.fill_between(x_data_all,conf_all_5,conf_all_95,color=cols[1],alpha=0.2,linewidth=1.,zorder=0)
if errb=="magnitude" or errb=="both":
cl1=ax.semilogx([x_data_all,x_data_all],[conf_all_5,conf_all_95],color=cols[1],linewidth=1.,zorder=1)
if errb=="return_time" or errb=="both":
cl2=ax.semilogx([conf_all_x_5,conf_all_x_95],[y_data_all,y_data_all],color=cols[1],linewidth=1.,zorder=1)
# Calculate GEV fit to data
shape,loc,scale=stats.genextreme.fit(data)
T=np.r_[1:10000]*0.01
# Perform K-S test and print goodness of fit parameters
D, p = stats.kstest(np.array(data).flatten(), 'genextreme', args=(shape, loc,scale));
print(plabel+' GEV fit, K-S test parameters p: '+str(p)+" D: "+str(D))
# Plot fit line
if dirn=="ascending":
rt=stats.genextreme.ppf(1./T,shape,loc=loc,scale=scale)
else:
rt=stats.genextreme.isf(1./T,shape,loc=loc,scale=scale)
l1=ax.semilogx(T,rt,color=cols[2],label="GEV fit")
# Highlight where the threshold is and where the return time curve bisects this threshold
xmin,xmax=ax.get_xlim()
ymin,ymax=ax.get_ylim()
ax.semilogx([xmin,xmax],[threshold,threshold],color="Grey",linestyle=tstyle,linewidth=2.5,label=tlabel, zorder=2)
nidx=find_nearest(y_data_all,threshold)
nidx1=np.where(rt>=threshold)[0][0]
ax.axvspan(conf_all_x_5[nidx],conf_all_x_95[nidx],ymin=0,ymax=(threshold-ymin)/(ymax-ymin),facecolor=cols[1],edgecolor=cols[2],linewidth=1.5,alpha=0.1,zorder=0)
#ax.axvline(x_data_all[nidx],ymin=0,ymax=(threshold-ymin)/(ymax-ymin),linestyle=tstyle,color=cols[0],linewidth=1.5,zorder=0)
ax.axvline(T[nidx1],ymin=0,ymax=(threshold-ymin)/(ymax-ymin),linestyle=tstyle,color=cols[0],linewidth=1.5,zorder=0)
# print(plabel+" return time with 5-95% range: "+str(x_data_all[nidx])+" ("+str(conf_all_x_5[nidx])+" "+str(conf_all_x_95[nidx])+")")
print(plabel+" return time with 5-95% range: "+str(T[nidx1])+" ("+str(conf_all_x_5[nidx])+" "+str(conf_all_x_95[nidx])+")")
# Return the fit parameters
return [shape,loc,scale]
fig = plt.figure()
fig.set_size_inches(21,16)
#ymin = 16,40, legend is lower right
plot_return_time('Control',' Max 2-day Precipitaiton, mm',p_twoday_max_c,f_twoday_max_c,30.04,"descending",fig,0,0)
plot_return_time('Double All','Max 2-day Precipitaiton, mm',p_twoday_max_da,f_twoday_max_da,30.04,"descending",fig,0,1)
plot_return_time('Quadruple All','Max 2-day Precipitaiton, mm',p_twoday_max_q,f_twoday_max_q,30.04,"descending",fig,0,2)
plot_return_time('Double Tropo','Max 2-day Precipitaiton, mm',p_twoday_max_dt,f_twoday_max_dt,30.04,"descending",fig,1,0)
plot_return_time('No Strat','Max 2-day Precipitaiton, mm',p_twoday_max_ns,f_twoday_max_ns,30.04,"descending",fig,1,1)
plt.savefig("Fig7/suk_djf_2day_max_pr_return_period_90th_percentile.png",dpi=600)
plt.show()