forked from Eddy-Stanford/QBO-History-Matching
-
Notifications
You must be signed in to change notification settings - Fork 0
/
analysis.py
265 lines (239 loc) · 8.99 KB
/
analysis.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
import argparse
import os
import joblib
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import xarray as xr
from history_matching.emulator import Emulator, chisquaredtest, implausibility2
from history_matching.samples import SampleSpace
from scipy.stats import sem
from sklearn.gaussian_process.kernels import RBF, ConstantKernel, WhiteKernel
import dispatch_utils
from qbo_utils.qbo_process import get_signal_period_amplitude
from qbo_utils.qbo_ref import fetch_qbo_file, get_reference_qbo
def get_qbo_period_amplitude(run_id, **config):
wave_base = dispatch_utils.get_wave_base_dir(**config)
path = os.path.join(
wave_base,
f"{str(run_id).zfill(2)}",
(
f"{str(run_id).zfill(2)}_"
f"QBO_{config['spinup']}"
f"_{config['time_to_run']}.nc"
),
)
with xr.open_dataset(path) as ds:
u = ds.ucomp.sel(pfull=10, method="nearest")
if config.get("verbose"):
fig, ax = plt.subplots()
ax.plot(u)
ax.set_xlabel("Time (days)")
ax.set_ylabel("u @ 10hPa (m/s)")
fig.savefig(os.path.join(wave_base, "analysis", f"u_{run_id}.png"))
plt.close(fig)
periods, amplitudes = get_signal_period_amplitude(
u,
smoothed_avg_months=config.get("months_to_smooth", 5),
points_per_month=30,
)
if config.get("verbose"):
print(f"Run:{run_id}, periods:{periods}")
print(f"Run:{run_id}, amplitudes:{amplitudes}")
return (
np.mean(periods),
sem(periods),
np.mean(amplitudes),
sem(amplitudes),
)
parser = argparse.ArgumentParser(description="Run analysis on HM wave")
parser.add_argument("configfile", type=str)
parser.add_argument("waveno", type=dispatch_utils.positive_int)
if __name__ == "__main__":
args = parser.parse_args()
config = dispatch_utils.load_config_file(args.configfile, args.waveno)
wave_base = dispatch_utils.get_wave_base_dir(**config)
exp_base = dispatch_utils.get_exp_base_dir(**config)
## Make directory for run analysis
os.makedirs(os.path.join(wave_base, "analysis"), exist_ok=True)
df = dispatch_utils.get_samples(**config)
## Get QBO files and convert them to numpy
## TODO: This code is pretty hardcoded rn, if we decide to
## look in to using more params/more observables this will need to
## be overhauled.
## TODO: Deal with this TODO
X_thisiter = np.zeros((config["nruns_per_wave"], 2))
y_thisiter = np.zeros((config["nruns_per_wave"], 2))
y_err_thisiter = np.zeros((config["nruns_per_wave"], 2))
for run, row in df.iterrows():
X_thisiter[run] = list(row)
period, period_err, amplitude, amplitude_err = get_qbo_period_amplitude(
run, **config
)
y_thisiter[run] = [period, amplitude]
y_err_thisiter[run] = [period_err, amplitude_err]
np.save(os.path.join(wave_base, "analysis", "X.npy"), X_thisiter)
np.save(os.path.join(wave_base, "analysis", "y.npy"), y_thisiter)
np.save(os.path.join(wave_base, "analysis", "y_err.npy"), y_err_thisiter)
# Suboptimal
X = np.concatenate(
[
*[
np.load(
os.path.join(
dispatch_utils.get_wave_base_dir(
name=config["name"], wave=p_wave
),
"analysis",
"X.npy",
)
)
for p_wave in range(args.waveno)
],
X_thisiter,
]
)
y = np.concatenate(
[
*[
np.load(
os.path.join(
dispatch_utils.get_wave_base_dir(
name=config["name"], wave=p_wave
),
"analysis",
"y.npy",
)
)
for p_wave in range(args.waveno)
],
y_thisiter,
]
)
y_err = np.concatenate(
[
*[
np.load(
os.path.join(
dispatch_utils.get_wave_base_dir(
name=config["name"], wave=p_wave
),
"analysis",
"y_err.npy",
)
)
for p_wave in range(args.waveno)
],
y_err_thisiter,
]
)
## Account for NaN values - just ignore
badys = np.isnan(y).any(axis=1)
X = X[~(badys)]
y = y[~(badys)]
y_err = y_err[~(badys)]
if config.get("verbose"):
print(f"y:{y} {y.shape}")
print(f"y_err:{y_err} {y_err.shape}")
period_ref, period_err_ref, amplitude_ref, amplitude_err_ref = get_reference_qbo(
fetch_qbo_file(local_path=os.path.join(exp_base, "qbo.dat"))
)
## EMULATOR
emulator = Emulator(
n_features=2,
kernel=ConstantKernel() * RBF(length_scale_bounds=(0.05, 1)) + WhiteKernel(),
random_state=config.get("random_state", 42),
n_restarts_optimizer=config.get("n_restarts_optimizer", 0),
)
emulator.fit(X, y, y_err=y_err)
if config.get("verbose"):
joblib.dump(emulator, os.path.join(wave_base, "analysis", "emulator.gpe"))
init_sample_space: SampleSpace = joblib.load(os.path.join(exp_base, "0.space"))
y_pred, y_pred_std = emulator.predict_over_space(
init_sample_space, return_std=True, resolution=SampleSpace.DEFAULT_RESOLUTION
)
# Calculate implausibility
imp_map = implausibility2(
y_pred,
y_pred_std,
np.array([period_ref, amplitude_ref]),
np.array([period_err_ref, amplitude_err_ref]),
)
df["implausibility"] = implausibility2(
xr.DataArray(y_thisiter.T),
xr.DataArray(y_err_thisiter.T),
np.array([period_ref, amplitude_ref]),
np.array([period_err_ref, amplitude_err_ref]),
).to_numpy()
df["period"] = y_thisiter.T[0]
df["amplitude"] = y_thisiter.T[1]
## Calculate New Non-implausible Space
# Use 3 sigma threshold
imp_cutoff = chisquaredtest(imp_map, config.get("significance_level", 0.01))
new_space = SampleSpace.from_xarray(imp_cutoff)
## Sample from new space
previous_sample_space = joblib.load(os.path.join(exp_base, f"{args.waveno}.space"))
new_space = new_space.intersection(previous_sample_space)
new_samples = new_space.uniform(config["nruns_per_wave"], labelled=True)
## Save samples and new space
joblib.dump(new_space, os.path.join(exp_base, f"{args.waveno + 1}.space"))
new_samples.to_csv(
os.path.join(exp_base, f"{args.waveno +1}_samples.csv"), index_label="run_id"
)
# Optional plotting code
if config.get("verbose"):
import matplotlib.patches as mpatches
fig, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=(16, 6))
CS = y_pred[0].plot.contour(ax=ax1, levels=25, x="cwtropics", y="Bt_eq")
ax1.scatter(X[:, 0], X[:, 1], marker="x")
ax1.clabel(CS, CS.levels, inline=True, fontsize=10)
ax1.set_ylabel("$Bt_{eq}$")
ax1.set_xlabel("$Cw_{tropics}$")
ax1.set_title("QBO Period (months)")
CS2 = y_pred[1].plot.contour(ax=ax2, levels=25, x="cwtropics", y="Bt_eq")
ax2.scatter(X[:, 0], X[:, 1], marker="x")
ax2.clabel(CS2, CS2.levels, inline=True, fontsize=10)
ax2.set_xlabel("$Cw_{tropics}$")
ax2.set_ylabel("$Bt_{eq}$")
ax2.set_title("QBO Amplitude (m/s)")
imp_map.plot.contour(ax=ax3, levels=25, x="cwtropics", y="Bt_eq")
CS3 = (
new_space.to_xarray()
.astype(int)
.interp(
cwtropics=np.linspace(
config["sample_space"]["cwtropics"]["min"],
config["sample_space"]["cwtropics"]["max"],
1000,
),
Bt_eq=np.linspace(
config["sample_space"]["Bt_eq"]["min"],
config["sample_space"]["Bt_eq"]["max"],
1000,
),
kwargs={"fill_value": 0},
)
.astype(bool)
.plot.contourf(
x="cwtropics",
y="Bt_eq",
ax=ax3,
alpha=0.3,
cmap="RdYlGn",
add_colorbar=False,
)
)
ax3.scatter(X[:, 0], X[:, 1], marker="x")
ax3.set_xlabel("$Cw_{tropics}$")
ax3.set_ylabel("$Bt_{eq}$")
ax3.legend(
handles=[
mpatches.Patch(
color="lightgreen", label="Next iteration non-implausible space"
),
mpatches.Patch(color="pink", label="Next iteration implausible space"),
]
)
ax3.set_title("Implausibility")
fig.savefig(os.path.join(wave_base, "analysis", "space.png"), dpi=300)
df.to_csv(os.path.join(wave_base, "analysis", "analysis.csv"))