-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathread_woa_csv.py
133 lines (104 loc) · 5.06 KB
/
read_woa_csv.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
import numpy as np
import pandas as pd
import tempfile
import gzip
import shutil
import wget
# ###############################################################
# ## Previous versions required data to be downloaded first to a folder
# wget -N -nH -nd -r -e robots=off --no-parent --force-html -A.nc https://data.nodc.noaa.gov/woa/WOA18/DATA/temperature/csv/decav/0.25/
# wget -N -nH -nd -r -e robots=off --no-parent --force-html -A.nc https://data.nodc.noaa.gov/woa/WOA18/DATA/salinity/csv/decav/0.25/
# ## Could also be done with bash script
# #!/bin/bash
# # t1='https://data.nodc.noaa.gov/woa/WOA18/DATA/temperature/csv/decav/0.25/woa18_decav_t'
# # t1_b="woa18_decav_t"
# # t3="an04.csv.gz"
# # for i in $(seq -f "%02g" 2 12);
# # do
# # wget $t1$i$t3
# # gunzip $t1_b$i$t3
# # done
def _woa_variable(variable):
_VAR = {
'temperature': 't',
'salinity': 's',
'silicate': 'i',
'phosphate': 'p',
'nitrate': 'n',
'oxygen_saturation': 'O',
'dissolved_oxygen': 'o',
'apparent_oxygen_utilization': 'A',
}
v = _VAR.get(variable)
if not v:
raise ValueError(
f'Unrecognizable variable. Expected one of {list(_VAR.keys())}, got "{variable}".'
)
return v
grids = {
'5': ('5deg', '5d'),
'1': ('1.00', '01'),
'1/4': ('0.25', '04'),
}
def read_woa(variable='temperature',resolution='1/4',field='an'):
print(field)
grid = grids.get(resolution)
woa_mat = np.zeros((len(lats_intervals),len(lons_intervals),len(woa_depths),len(month_intervals)))
for i in np.arange(12):
print('Reading month',i)
## Create the woa url to download
baseurl='https://data.nodc.noaa.gov/woa/WOA18/DATA/'
baseurl=f'{baseurl}{variable}/csv/decav/{grid[0]}/'
suf = 'woa18_decav_'
woa_file = '{suf}{var}{foo:02d}{field}{ress}.csv'.format(foo=i,var= _woa_variable(variable),suf=suf,pref=pref,field=field,ress=grid[1])
woa_url = f'{baseurl}{woa_file}.gz'
## Download de WOA file as a temporary file
with tempfile.TemporaryDirectory() as tmpdirname:
print('created temporary directory', tmpdirname)
ftemp=f'{tmpdirname}/WOA.gz'
print('Beginning file download with urllib2...')
wget.download(woa_url, ftemp)
print('downloaded to ',ftemp)
ftemp_csv=f'{tmpdirname}/WOA.csv'
## uncompress the gzip csv file
with gzip.open(ftemp, 'rb') as f_in:
with open(ftemp_csv, 'wb') as f_out:
shutil.copyfileobj(f_in, f_out)
## Read and reshape csv file
woa_s = pd.read_csv(ftemp_csv,skiprows= 2,header=None,na_values='',names=['lat','lon'] +woa_depths)
woa_s['lat_bins'] = pd.cut(woa_s['lat'],bins=lats_intervals)
woa_s['lon_bins'] = pd.cut(woa_s['lon'],bins=lons_intervals)
res= woa_s.groupby(['lat_bins','lon_bins']).mean()
##res.drop(['lat','lon'],inplace=True)
res_np = res.values
res_np = res_np[:,2:]
# mat_data = np.reshape(res[5].values,(len(lats_intervals),len(lons_intervals)))
woa_mat[:,:,:,i] = np.reshape(res_np,(len(lats_intervals),len(lons_intervals),len(woa_depths)))
return woa_mat
##############################################
## Recreate WOA intervals for binning
woa_depths = [0,5,10,15,20,25,30,35,40,45,50,55,60,65,70,75,80,85,90,95,100,125,150,175,200,225,250,275,300,325,350,375,400,425,450,475,500,550,600,650,700,750,800,850,900,950,1000,1050,1100,1150,1200,1250,1300,1350,1400,1450,1500]
woa_depths_lefts = [-5,0,5,10,15,20,25,30,35,40,45,50,55,60,65,70,75,80,85,90,95,100,125,150,175,200,225,250,275,300,325,350,375,400,425,450,475,500,550,600,650,700,750,800,850,900,950,1000,1050,1100,1150,1200,1250,1300,1350,1400,1450]
woa_depths_rights = [5,10,15,20,25,30,35,40,45,50,55,60,65,70,75,80,85,90,95,100,125,150,175,200,225,250,275,300,325,350,375,400,425,450,475,500,550,600,650,700,750,800,850,900,950,1000,1050,1100,1150,1200,1250,1300,1350,1400,1450,1500,1550]
depth_intervals = pd.IntervalIndex.from_arrays((np.array(woa_depths)+np.array(woa_depths_lefts))/2.,(np.array(woa_depths)+np.array(woa_depths_rights))/2.)
month_intervals = pd.interval_range(start=0.5, end=12.5,freq=1)
##############################################
## Lat lon bins depend on grid size
grid_size = 0.25
lats_intervals = pd.interval_range(start=-90, end=90,freq=grid_size)
lons_intervals = pd.interval_range(start=-180, end=180,freq=grid_size)
temp_mat = read_woa(variable='temperature',resolution='1/4',field='an'
)
sal_mat = read_woa(suf = './data/woa18_decav_s',
pref = 'an04.csv')
## Roll axis so it has same dimensions as netcdf
temp_mat=np.rollaxis(temp_mat,3,0)
temp_mat=np.rollaxis(temp_mat,3,1)
## Roll axis so it has same dimensions as netcdf
sal_mat=np.rollaxis(sal_mat,3,0)
sal_mat=np.rollaxis(sal_mat,3,1)
np.savez('data/woa_sal_04.npz',sal_mat)
np.savez('data/woa_temp_04.npz',temp_mat)
## How to read
sal_mat = np.load('data/woa_sal_04.npz')['arr_0']
temp_mat = np.load('data/woa_temp_04.npz')['arr_0']