-
Notifications
You must be signed in to change notification settings - Fork 1
/
parq_to_nc.py
233 lines (204 loc) · 9.07 KB
/
parq_to_nc.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
"""
creates the actual ESACCI dataset from the parquet files produces by the ML pipeline
"""
import pandas as pd
import netCDF4 as nc4
import os
import glob
import time
import warnings
import traceback
import numpy as np
from tqdm import tqdm
import multiprocessing as mlp
from datetime import datetime
import sys
ctnames = [ "Ci", "As", "Ac", "St", "Sc", "Cu", "Ns", "Dc"]
propnames = ['twp', 'lwp', 'iwp', 'cerl', 'ceri', 'cod', 'ptop', 'tsurf','clt']
#"rlut", "rsut", "rsutcs", "rlutcs"]
units = ["kg/m2","kg/m2","kg/m2","m","m","1","Pa","K","1"]#,"1","1","1","1"]
clear = ["clear"]
s_names = ["undetermined_cloud_type_or_clear_sky",
"cirrus_or_cirrostratus", "altostratus", "altocumulus",
"stratocumulus", "stratus", "cumulus",
"nimbostratus", "deep_convective",
"atmosphere_mass_content_of_cloud_condensed_water", "atmosphere_mass_content_of_cloud_liquid_water", "atmosphere_mass_content_of_cloud_ice",
"effective_radius_of_cloud_liquid_water_particles",
"effective_radius_of_cloud_ice_particles",
"atmosphere_optical_thickness_due_to_cloud",
"air_pressure_at_cloud_top",
"surface_temperature",
"cloud_area_fraction",
"toa_upwelling_longwave_flux",
"toa_upwelling_shortwave_flux",
"toa_upwelling_shortwave_flux_assuming_clear_sky",
"toa_upwelling_longwave_flux_assuming_clear_sky"]
l_names = ["predicted relative frequency of occurence of the 'undetermined' type",
"predicted relative frequency of occurence of the 'Ci' type",
"predicted relative frequency of occurence of the 'As' type",
"predicted relative frequency of occurence of the 'Ac' type",
"predicted relative frequency of occurence of the 'St' type",
"predicted relative frequency of occurence of the 'Sc' type",
"predicted relative frequency of occurence of the 'Cu' type",
"predicted relative frequency of occurence of the 'Ns' type",
"predicted relative frequency of occurence of the 'Dc' type",
"vertically integrated cloud water",
"vertically integrated liquid cloud water",
"vertically integrated cloud ice",
"effective radius of cloud liquid droplets",
"effective radius of cloud ice particles",
"cloud optical thickness",
"cloud top pressure",
"surface temperature", "2D cloud cover",
"all-sky top of atmosphere longwave upwelling solar radiation",
"all-sky top of atmosphere shortwave upwelling solar radiation",
"top of atmosphere shortwave upwelling solar radiation assuming clear sky",
"top of atmosphere longwave upwelling solar radiation assuming clear sky"]
def dayproc(data,d,day):
"""computes the daily 1°x1° averages and assigns them to the correct boxes in the output
Args:
data (pd.Dataframe): CCClim dataframe with up to 2 daily values
d (int): the dth nonezero entry
day (daiy): calendar day
Returns:
tuple: (index, properly managed daily averages)
"""
data = data.groupby(["lat","lon","time"]).mean().reset_index()
temp = data.set_index("time")
temp = temp.loc[day]
temp = temp.set_index(["lat","lon"]).unstack()
out = []
for c,col in enumerate(clear+ctnames+propnames):
vals = temp.loc[:,col].values
#this handles if there are missing grid boxes
if not np.all(vals.shape==(180,360)):
nanframe=pd.DataFrame(index=np.arange(-90,90,1.),columns=np.arange(-180,180,1.))
nanframe.update(temp.loc[:,col])
vals = nanframe.values
out.append(vals)
out = np.stack(out,axis=-1)
return d,out
def assign(ds,vals,shift,j,prop):
"""puts the data in the nc file with correct name and units
Args:
ds (nc4.Dataset): ouput dataset
vals (np.ndarray): output data
shift (int): units and names and stuff dont all have the same length
j (int): index of property
prop (string): name of property
"""
shifted_j = j+shift
arr = vals[...,shifted_j]
temp = ds.createVariable(prop,"f4",("time","lat","lon",))
temp.standard_name = s_names[shifted_j]
temp.long_name = l_names[shifted_j]
if prop in propnames:
temp.units = units[j]
if units[j]=="kg/m2":
arr/=1000
elif units[j]=="Pa":
arr*=100
elif units[j]=="m":
arr/=1e6
else:
temp.units= "1"
temp[:] = arr
def modname(inname):
"""modifies the detailed input data name to the effective output data"""
return inname.replace("parquet", "nc").replace(
"ESACCI_de","CCClim_").replace("frame100_10000_10_0123459","")
class Dayproc(object):
"""ensures that I can use dayproc with variable ancillary inputs in subprocesses"""
def __init__(self, df):
self.df = df
def __call__(self, x,y):
return dayproc( self.df,x,y)
def exists(path,outfolder):
"""shorthand to check wether the CCClim file for the day already exists"""
bname = os.path.basename(path)
outpath = os.path.join(outfolder,modname(bname))
return os.path.exists(outpath)
def main(year=None):
"""easy way to call the pipeline for each individual year
Args:
year (int, optional): If not specified does everything, which is possible I guess but takes ages . Defaults to None.
"""
scratch = os.environ["SCR"]
infolder = os.path.join(scratch, "ESACCI/parquets/w_anc")
outfolder = os.path.join(scratch,"ESACCI","CCClim")
if not os.path.exists(outfolder):
os.makedirs(outfolder)
files = glob.glob(os.path.join(infolder,"ESACCI_de????f*.parquet"))[:]
files = [x for x in files if not exists(x,outfolder)]
if year is not None:
files = [x for x in files if year in x]
assert len(files)>0
for file in tqdm(files,total=len(files)):
name = os.path.basename(file)
outds = nc4.Dataset(os.path.join(outfolder,modname(name)),
"w",format = "NETCDF4")
time_out = outds.createDimension("time", None)
lat_out = outds.createDimension("lat", 180)
lon_out = outds.createDimension("lon", 360)
outds.title = "CCClim"
outds.comment = "Machine learning powered Cloud Class Climatology based on ESA Cloud_cci"
outds.Conventions = "CF-1.10"
outds.institution = "Deutsches Zentrum für Luft und Raumfahrt - Institut für Physik der Atmosphäre"
outds.source = "satellite observations"
times_out = outds.createVariable("time","f4",("time",))
lats_out = outds.createVariable("lat","f4",("lat",))
lons_out = outds.createVariable("lon","f4",("lon",))
times_out.units = "days since 1970-01-01"
lats_out.units = "degrees_north"
lons_out.units = "degrees_east"
times_out.standard_name = "time"
lons_out.standard_name = "longitude"
lats_out.standard_name = "latitude"
lats_out[:]=np.arange(-90,90,1.)
lons_out[:] = np.arange(-180,180,1.)
data = pd.read_parquet(file)
if "2010" in file:
#exclude july 2010 because that has wrong data
start = datetime.fromisoformat("1970-01-01")
start_july = datetime.fromisoformat("2010-07-01")
end_july = datetime.fromisoformat("2010-07-31")
ex_july = (data.time>(end_july-start).days)|( data.time<(start_july-start).days)
data=data[ex_july]
elif "1994" in file:
#exclude last quarter of 1994 because weird/missing data
start = datetime.fromisoformat("1970-01-01")
start_january = datetime.fromisoformat("1994-01-01")
ex_260 = data.time- (start_january-start).days < 260
data=data[ex_260]
elif "1986" in file:
#exclude first january
start = datetime.fromisoformat("1970-01-01")
start_january = datetime.fromisoformat("1986-01-01")
ex_11 = data.time!=(start_january-start).days
data=data[ex_11]
un_time = data.time.unique()
un_time.sort()
times_out[:] = un_time
pool=mlp.Pool(10)
# I dont know why i carry the enumerate index around here but there must be a reason
outs = pool.starmap(Dayproc(data),list(enumerate(un_time)))
while True:
#you could also do this i tink: out= [x for x in out if x is not None]
try:
outs.remove(None)
except Exception:
break
outs.sort()
base=np.stack([x for _,x in outs])
print(base.shape)
for j, prop in enumerate(propnames):
assign(outds,base,len(clear+ctnames),j,prop)
for j,cname in enumerate(clear+ctnames):
assign(outds,base,0,j,cname)
outds.close()
if __name__=="__main__":
try:
year=sys.argv[1]
except IndexError:
year=None
main(year = year)