-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy path3_gfs_to_rank.py
169 lines (151 loc) · 6.03 KB
/
3_gfs_to_rank.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
# -*- coding: utf-8 -*-
import numpy as np
import pandas as pd
import scipy.stats as stats
#import matplotlib.pyplot as plt
import os,sys
from datetime import datetime
import COMMON as COM
##################################################
## 抽出地点と気象変数の指定
ENCODE = "cp932"
SDP_LIST = pd.read_csv(COM.INFO_PATH +"/"+ "sdp_list.csv",index_col="SDP",encoding=ENCODE)
## ランキング計算の入力
# 以下のSTAT,DATA
STAT_PATH = COM.HCST_PATH #"./hindcast"
DATA_PATH = COM.FCST_PATH #"./forecast"
## ランキング結果の出力先
OUT_PATH = COM.INFO_PATH #"./info"
print("enter:", sys.argv)
print("now:", datetime.now())
print("in1:", STAT_PATH)
print("in2:", DATA_PATH)
print("out:", OUT_PATH)
os.makedirs(OUT_PATH, exist_ok=True)
## 気象変数の単位
GFS_LIST = pd.read_csv(COM.INFO_PATH +"/"+ "gfs_list.csv",index_col="GFS")
UNITS = {}
ABBREVIATION = {}
LONG_NAME = {}
for v in GFS_LIST.index:
UNITS[v] = GFS_LIST.loc[v,"units"]
ABBREVIATION[v] = GFS_LIST.loc[v,"abbreviation"]
LONG_NAME[v] = GFS_LIST.loc[v,"long_name"]
##################################################
## 統計値の読み込み
STAT = pd.read_csv(DATA_PATH +"/"+ "gfs_mean1d.csv")
FPCT = {}
FKDE = {}
for v in STAT.columns[1:]:
# パーセンタイル値: p0,p1,...,p99,p100
PCT = STAT.loc[3:103,v].values
# PCT関数: GFS変数→パーセンタイル
FPCT[v] = np.vectorize(lambda x,PCT=np.copy(PCT): np.searchsorted(PCT, x)/1.01)
# KDE関数: GFS変数→確率
n = STAT.loc[0,v]
sig = STAT.loc[2,v]
iqr = PCT[75] - PCT[25]
h = 0.9*min(sig,iqr/1.34)/(n**0.2) # optimal bandwidth for KDE
if h == 0.:
FKDE[v] = lambda x: 1.0
else:
FKDE[v] = np.vectorize(lambda x,PCT=np.copy(PCT),h=h: sig*np.mean(stats.norm.pdf(x,PCT,h)))
# デバッグ用: KDE可視化
"""
if v in ['v-component_of_wind_tropopause_00',]:
import matplotlib.pyplot as plt
x = np.linspace(PCT[0]-sig,PCT[100]+sig,100)
y = FKDE[v](x)
plt.figure()
plt.title(v)
plt.xlabel(v)
plt.plot(x,y)
plt.plot(PCT,np.linspace(0,0,101),marker="o")
plt.grid()
plt.show()
"""
##################################################
## ランキングの計算処理
RANK = pd.DataFrame([],columns=["SDP","FUKEN","NAME","DATE","GFS","PCTL","LOGP","MEAN"])
## ランク計算パラメータ
P01,P99,KDE,EPS = 1.,99.,1e-2,1e-300
SKIP = {
'Pressure_convective_cloud_bottom',
'Pressure_convective_cloud_top',
'Pressure_high_cloud_bottom_Mixed_intervals_Average',
'Pressure_high_cloud_top_Mixed_intervals_Average',
'Pressure_low_cloud_bottom_Mixed_intervals_Average',
'Pressure_low_cloud_top_Mixed_intervals_Average',
'Pressure_middle_cloud_bottom_Mixed_intervals_Average',
'Pressure_middle_cloud_top_Mixed_intervals_Average',
}
#SDP_LIST = SDP_LIST[20:25]
for SDP in SDP_LIST.index[:]:
NAME = SDP_LIST.loc[SDP,'NAME']
FUKEN = SDP_LIST.loc[SDP,'FUKEN']
##################################################
## 予報値
DATA = pd.read_csv("%s/%05d.csv"%(DATA_PATH,SDP),parse_dates=[0],index_col=0)
DATA = DATA[1:] # 1日8コマに整列（開始21時を捨てる）
DATA = DATA.resample("1D").mean()
for v in DATA.columns[1:]:
if not(v in STAT.columns[1:]): continue
if v[:-3] in SKIP: continue
VAL = DATA[v]
# パーセンタイル値のスコア
PCT = pd.Series(FPCT[v](VAL), index=DATA.index)
# 確率密度分布関数のスコア
LOG = pd.Series(-np.log10(FKDE[v](VAL)+EPS), index=DATA.index)
# 両スコアからレア日を列挙
COND = ((PCT<P01) | (PCT>P99)) & (LOG>-np.log10(KDE))
for d in COND[COND].index:
print(sys.argv[0], SDP,NAME,d,v,PCT[d],LOG[d],VAL[d])
row = pd.Series([SDP,FUKEN,NAME,d,v,PCT[d],LOG[d],VAL[d]],index=RANK.columns)
RANK = RANK.append(row, ignore_index=True)
##################################################
"""
## 大小に振れる変数を除外
VARS = RANK.groupby(["GFS"]).mean()["PERCENTILE"]
VARS = VARS[(VARS<P01) | (VARS>P99)]
RANK = RANK[[v in list(VARS.index) for v in RANK.GFS]]
"""
## 事象リストの保存
RANK["units"] = RANK.apply(lambda x: UNITS[x["GFS"][:-3]],axis=1)
RANK["abbreviation"] = RANK.apply(lambda x: ABBREVIATION[x["GFS"][:-3]],axis=1)
RANK["long_name"] = RANK.apply(lambda x: LONG_NAME[x["GFS"][:-3]],axis=1)
RANK = RANK.sort_values(["DATE","SDP","GFS"])
RANK = RANK[["DATE","SDP","FUKEN","NAME","abbreviation","MEAN","units","PCTL","long_name"]]
RANK.to_csv(OUT_PATH +"/"+ "gfs_rank.csv",encoding=ENCODE)
#sys.exit(0)
##################################################
"""
## 集約ランキングの保存
SCORE = "SCORE"
RANK = RANK.rename(columns={"NAME":SCORE})
## 変数ランキング
vRANK = RANK.groupby(["GFS"]).agg({SCORE:'count','PERCENTILE':'mean','AVG':'mean','SDP':lambda x:set(x),'DATE':lambda x:set(x)})
vRANK = vRANK.sort_values([SCORE],ascending=[False])
vRANK = vRANK.reset_index()
vRANK["units"] = vRANK.apply(lambda x: UNITS[x["GFS"][:-3]],axis=1)
vRANK = vRANK[["GFS","SCORE","PERCENTILE","AVG","units","SDP","DATE"]]
vRANK = vRANK.rename(columns={"SDP":"#SDP"})
vRANK.to_csv(OUT_PATH +"/"+ "var_rank.csv",encoding=ENCODE)
## 地点ランキング
sRANK = RANK.groupby(["SDP"]).agg({SCORE:'count','DATE':lambda x:set(x),'GFS':lambda x:set(x)})
sRANK = sRANK.sort_values([SCORE],ascending=[False])
sRANK = sRANK.join(SDP_LIST[["FUKEN","NAME"]])
sRANK = sRANK.reset_index()
sRANK = sRANK[["SDP","SCORE","NAME","DATE","GFS"]]
sRANK = sRANK.rename(columns={"GFS":"#GFS"})
sRANK.to_csv(OUT_PATH +"/"+ "sdp_rank.csv",encoding=ENCODE)
## 日付ランキング
dRANK = RANK.groupby(["DATE"]).agg({SCORE:'count','SDP':lambda x:set(x),'GFS':lambda x:set(x)})
dRANK = dRANK.sort_values([SCORE],ascending=[False])
dRANK = dRANK.reset_index()
dRANK = dRANK[["DATE","SCORE","SDP","GFS"]]
dRANK = dRANK.rename(columns={"SDP":"#SDP","GFS":"#GFS"})
dRANK.to_csv(OUT_PATH +"/"+ "day_rank.csv",encoding=ENCODE)
"""
##################################################
print("leave:", sys.argv)
sys.exit(0)