def rmse(root, index):
times = [
datetime.utcfromtimestamp(int(t)) for t in nc.getvar(root, 'time')[:]
]
days = [t.date() for t in times]
days.sort()
days_index = [d.day for d in set(days)]
days_amount = len(days_index)
nc.getdim(root, 'diarying', days_amount)
nc.sync(root)
measurements = nc.getvar(root, 'measurements')
estimated = nc.getvar(root, 'globalradiation')
error_diff = nc.getvar(root, 'errordiff', 'f4', (
'time',
'yc_cut',
'xc_cut',
), 4)
error = nc.getvar(root, 'error', 'f4', (
'time',
'yc_cut',
'xc_cut',
), 4)
diary_error = nc.getvar(root, 'diaryerror', 'f4', (
'diarying',
'yc_cut',
'xc_cut',
), 4)
error_diff[:] = np.zeros(estimated.shape)
error[:] = np.zeros(estimated.shape)
diary_error[:] = np.zeros(
(days_amount, estimated.shape[1], estimated.shape[2]))
nc.sync(root)
#the_max = measurements[:].max()
error_diff[:,
index, :] = measurements[:, index, :] - estimated[:, index, :]
error[:, index, :] = np.abs(error_diff[:, index, :])
nc.sync(root)
max_value_in_day = np.zeros([days_amount]) + 1
for i in range(len(days)):
d_i = days_index.index(days[i].day)
max_value_in_day[d_i] = measurements[
i, index, 0] if max_value_in_day[d_i] < measurements[
i, index, 0] else max_value_in_day[d_i]
diary_error[d_i, index, :] += np.array([error_diff[i, index, 0]**2, 1])
count = diary_error[:, index, 1]
count[count == 0] = 1
diary_error[:, index, 0] = np.sqrt(diary_error[:, index, 0] / count)
diary_error[:, index,
1] = diary_error[:, index, 0] / max_value_in_day * 100
show("\rDiary RMS error: %.2f" % (diary_error[:, index, 1]).mean())
for i in range(len(days)):
d_i = days_index.index(days[i].day)
error[i, index, 1] = error[i, index, 1] / max_value_in_day[d_i] * 100
result = np.sum(error[:, index, 1]**2)
result = np.sqrt(result / error.shape[0])
show("Half-hour RMS error: %.2f \n" % result)
#diary_error[:, index,1] = diary_error[:, index,0]
nc.sync(root)
nc.close(root)
def cut_projected(root):
lat = nc.getvar(root, 'lat')
lon = nc.getvar(root, 'lon')
time = nc.getvar(root, 'data_time')
months = list(set([ (datetime.fromtimestamp(int(t))).month for t in time ]))
nc.getdim(root, 'monthing')
months_cut = nc.getvar(root, 'months', 'i2', ('monthing',))
dims = list(lat.dimensions)
dims.insert(0, 'monthing')
linke = nc.getvar(root, 'linketurbidity', 'f4', tuple(dims),4)
linke_x, linke_y = project_coordinates(lat[:], lon[:])
months_cut[:] = np.array(list(months))
for i in range(len(months)):
linke[i] = cut_month(linke_x, linke_y, months[i])
return linke[:], linke_x, linke_y
def rows2netcdf(rows, filename, index):
root, is_new = nc.open(filename)
if not is_new:
measurements = nc.getvar(root, 'measurements', 'f4', ('timing','northing','easting'), 4)
slots = nc.getvar(root, 'slots')
times = [ datetime.fromtimestamp(int(t)) for t in nc.getvar(root, 'time') ]
instant_radiation = rows2slots(rows,2)
i_e = 0
i_m = 0
while i_e < len(times) and i_m < len(instant_radiation):
#print i_e,"/", len(times) -1, " | ", i_m, "/", len(instant_radiation) - 1
# When date estimated before date measured
if times[i_e].date() < instant_radiation[i_m][1][0].date():
i_e += 1
# When date estimated after date measured
elif times[i_e].date() > instant_radiation[i_m][1][0].date():
i_m += 1
else:
if slots[i_e] < instant_radiation[i_m][0]:
measurements[i_e, index,:] = np.array([-1, -1])
i_e += 1
elif slots[i_e] > instant_radiation[i_m][0]:
i_m += 1
else:
value = instant_radiation[i_m][1][1]
row_in_slot = instant_radiation[i_m][1][2]
measurements[i_e, index,:] = np.array([value, value])
i_e += 1
i_m += 1
days = [ t.date() for t in times ]
nc.getdim(root, 'diarying')
diary_error = nc.getvar(root, 'diaryerror', 'f4', ('diarying', 'northing', 'easting'))
estimated = nc.getvar(root, 'globalradiation')[:]
error_diff = nc.getvar(root, 'errordiff', 'f4', ('timing','northing','easting'), 4)
error = nc.getvar(root, 'error', 'f4', ('timing','northing','easting'), 4)
error_diff[:, index, :] = measurements[:,index,:] - estimated[:,index,:]
error[:, index, :] = np.sqrt((error_diff[:, index, :])**2)
print error_diff.shape
nc.sync(root)
for i in range(len(days)):
d = days[i].day
print index, i, d, error_diff[i,index], diary_error[d, index]
diary_error[d, index,:] += error_diff[i,index,:]
diary_error[:, index,:] = np.sqrt((diary_error[:, index,:])**2)
nc.close(root)
def merge_tails():
import glob
import numpy as np
#from scipy import misc
say("Merging all the tails in one netCDF file...")
root, is_new = nc.open(localpath+ "/dem.nc")
n_d = nc.getdim(root,'northing', 180 * 120) # 180 degrees * 120 px/degree
n_e = nc.getdim(root,'easting', 360 * 120) # 360 degrees * 120 px/degree
data = nc.getvar(root, 'dem', 'i2', ('northing','easting',))
files = glob.glob(localpath + "/all10/all10/*10g")
tails = [[file_coordinates(f),f] for f in files]
for t, f in tails:
print t, f
tmp = np.fromfile(f,np.int16)
dynamic = tmp.shape[0]/10800
tmp = tmp.reshape(dynamic,10800)
#misc.imsave(f+'.png',tmp)
data[t[0]:t[0]+dynamic,t[1]:t[1]+10800] = tmp
nc.close(root)
def do_file(self, filename, data):
print "Creating ", filename
# create compact file and initialize basic settings
root, is_new = nc.open(filename + self.extension) # The filename does not contain the extension
if is_new:
sample, n = nc.open(data[(data.keys())[0].completepath()])
shape = sample.variables['data'].shape
nc.getdim(root,'northing', shape[1])
nc.getdim(root,'easting', shape[2])
nc.getdim(root,'timing')
v_lat = nc.getvar(root,'lat', 'f4', ('northing','easting',), 4)
v_lon = nc.getvar(root,'lon', 'f4', ('northing','easting',), 4)
v_lon[:] = nc.getvar(sample, 'lon')[:]
v_lat[:] = nc.getvar(sample, 'lat')[:]
nc.close(sample)
if isinstance(data,collections.Mapping):
for d in data:
self.do_var(root, d, data[d])
else:
self.create_var(root, 'data', data)
# save the content inside the compact file
if not root is None: nc.close(root)
f = File(localname=localname, downloaded=True, begin_download=begin_time, end_download=datetime.now())
f.save()
return f
def rmse(root, index):
times = [ datetime.utcfromtimestamp(int(t)) for t in nc.getvar(root, 'data_time')[:] ]
days = [ t.date() for t in times ]
days.sort()
days_index = [d.day for d in set(days)]
days_amount = len(days_index)
nc.getdim(root, 'diarying', days_amount)
nc.sync(root)
measurements = nc.getvar(root, 'measurements')
estimated = nc.getvar(root, 'globalradiation')
error_diff = nc.getvar(root, 'errordiff', 'f4', ('timing','northing_cut','easting_cut',), 4)
error = nc.getvar(root, 'error', 'f4', ('timing','northing_cut','easting_cut',), 4)
diary_error = nc.getvar(root, 'diaryerror', 'f4', ('diarying', 'northing_cut', 'easting_cut',), 4)
error_diff[:] = np.zeros(estimated.shape)
error[:] = np.zeros(estimated.shape)
diary_error[:] = np.zeros((days_amount, estimated.shape[1], estimated.shape[2]))
nc.sync(root)
#the_max = measurements[:].max()
error_diff[:, index, :] = measurements[:,index,:] - estimated[:,index,:]
error[:, index, :] = np.abs(error_diff[:, index, :])
nc.sync(root)
max_value_in_day = np.zeros([days_amount]) + 1
for i in range(len(days)):
d_i = days_index.index(days[i].day)
max_value_in_day[d_i] = measurements[i,index,0] if max_value_in_day[d_i] < measurements[i,index,0] else max_value_in_day[d_i]
diary_error[d_i, index,:] += np.array([ error_diff[i,index,0] ** 2,1])
count = diary_error[:, index, 1]
count[count == 0] = 1
diary_error[:, index,0] = np.sqrt(diary_error[:, index, 0] / count)
diary_error[:, index,1] = diary_error[:, index,0] / max_value_in_day * 100
show("\rDiary RMS error: %.2f" % (diary_error[:, index, 1]).mean())
for i in range(len(days)):
d_i = days_index.index(days[i].day)
error[i,index,1] = error[i,index,1] / max_value_in_day[d_i] * 100
result = np.sum(error[:, index, 1] ** 2)
result = np.sqrt(result / error.shape[0])
show("Half-hour RMS error: %.2f \n" % result)
#diary_error[:, index,1] = diary_error[:, index,0]
nc.sync(root)
def cut_projected(root):
lat = nc.getvar(root, "lat")
lon = nc.getvar(root, "lon")
time = nc.getvar(root, "time")
months = list(set([(datetime.fromtimestamp(int(t))).month for t in time]))
months_d = nc.getdim(root, "monthing")
months_cut = nc.getvar(root, "months", "i2", ("monthing",))
linke = nc.getvar(root, "linketurbidity", "f4", ("monthing", "northing", "easting"), 4)
linke_x, linke_y = project_coordinates(lat[:], lon[:])
months_cut[:] = np.array(list(months))
for i in range(len(months)):
linke[i] = cut_month(linke_x, linke_y, months[i])
return linke[:], linke_x, linke_y
def do(self, stream):
from libs.sat.goes import calibration
resultant_stream = stream.clone()
for fs in stream.files.all():
f = fs.file
if f.channel() == '01':
sat = f.satellite()
root, is_new = nc.open(f.completepath())
nc.getdim(root,'coefficient',1)
var = nc.getvar(root, 'counts_shift', 'f4', ('coefficient',), 4)
var[0] = calibration.counts_shift.coefficient(sat)
var = nc.getvar(root, 'space_measurement', 'f4', ('coefficient',), 4)
var[0] = calibration.space_measurement.coefficient(sat)
var = nc.getvar(root, 'prelaunch', 'f4', ('coefficient',), 4)
var[0] = calibration.prelaunch.coefficient(sat)[0]
var = nc.getvar(root, 'postlaunch', 'f4', ('coefficient',), 4)
dt = f.datetime()
var[0] = calibration.postlaunch.coefficient(sat, dt.year, dt.month)
#data = np.float32(self.calibrated_coefficient) * ((data / np.float32(self.counts_shift)) - np.float32(self.space_measurement))
nc.close(root)
fs.processed=True
fs.save()
return resultant_stream
def do(self, stream):
from libs.sat.goes import calibration
resultant_stream = stream.clone()
for ms in stream.materials.all():
m = ms.material
if hasattr(m, 'channel') and hasattr(m, 'satellite') and m.channel() == '01':
sat = m.satellite()
root = nc.open(m.completepath())[0]
nc.getdim(root,'coefficient',1)
var = nc.getvar(root, 'counts_shift', 'f4', ('coefficient',), 4)
var[0] = calibration.counts_shift.coefficient(sat)
var = nc.getvar(root, 'space_measurement', 'f4', ('coefficient',), 4)
var[0] = calibration.space_measurement.coefficient(sat)
var = nc.getvar(root, 'prelaunch', 'f4', ('coefficient',), 4)
var[0] = calibration.prelaunch.coefficient(sat)[0]
var = nc.getvar(root, 'postlaunch', 'f4', ('coefficient',), 4)
dt = m.datetime()
var[0] = calibration.postlaunch.coefficient(sat, dt.year, dt.month)
#data = np.float32(self.calibrated_coefficient) * ((data / np.float32(self.counts_shift)) - np.float32(self.space_measurement))
nc.close(root)
ms.processed=True
ms.save()
return resultant_stream
def do_file(self, filename, stream):
# create compact file and initialize basic settings
begin_time = self.getdatetimenow()
root, is_new = nc.open(filename)
if is_new:
sample, n = nc.open(stream.files.all()[0].file.completepath())
shape = sample.variables['data'].shape
nc.getdim(root,'northing', shape[1])
nc.getdim(root,'easting', shape[2])
nc.getdim(root,'timing')
v_lat = nc.getvar(root,'lat', 'f4', ('northing','easting',), 4)
v_lon = nc.getvar(root,'lon', 'f4', ('northing','easting',), 4)
v_lon[:] = nc.getvar(sample, 'lon')[:]
v_lat[:] = nc.getvar(sample, 'lat')[:]
nc.close(sample)
nc.sync(root)
self.do_var(root, 'data', stream)
# save the content inside the compact file
if not root is None: nc.close(root)
f = File(localname=filename)
f.save()
return f
def dailyerrors(root, stations):
times = [
datetime.fromtimestamp(int(t)) for t in nc.getvar(root, 'time')[:]
]
days = [t.date() for t in times]
days.sort()
days_index = [d.day for d in set(days)]
days_index.sort()
days_amount = len(days_index)
nc.getdim(root, 'diarying', days_amount)
nc.sync(root)
measurements = nc.getvar(root, 'measurements')
estimated = nc.getvar(root, 'globalradiation')
error_diff = root.getvar('errordiff', 'f4', (
'time',
'yc_cut',
'xc_cut',
), 4)
RMS_daily_error = root.getvar('RMSdailyerror', 'f4', (
'diarying',
'yc_cut',
'xc_cut',
), 4)
BIAS_daily_error = root.getvar('BIASdailyerror', 'f4', (
'diarying',
'yc_cut',
'xc_cut',
), 4)
error_diff[:] = np.zeros(estimated.shape)
RMS_daily_error[:] = np.zeros(
(days_amount, estimated.shape[1], estimated.shape[2]))
BIAS_daily_error[:] = np.zeros(
(days_amount, estimated.shape[1], estimated.shape[2]))
nc.sync(root)
for s in stations:
index = stations.index(s)
show('Station: %s \n' % stations[index])
error_diff[:,
index, :] = measurements[:, index, :] - estimated[:,
index, :]
nc.sync(root)
sum_value_in_day = np.zeros((days_amount))
for i in range(len(days)):
d_i = days_index.index(days[i].day)
if not measurements[i, index, 0] == 0.0:
sum_value_in_day[d_i] += measurements[i, index, 0]
RMS_daily_error[d_i, index, :] += np.array(
[error_diff[i, index, 0]**2, 1])
BIAS_daily_error[d_i, index, :] += error_diff[i, index, 0]
count = RMS_daily_error[:, index, 1]
count[count == 0] = 1
RMS_daily_error[:, index, 0] = np.sqrt(
RMS_daily_error[:, index, 0] / count) / sum_value_in_day * 100
BIAS_daily_error[:, index, 0] = (BIAS_daily_error[:, index, 0] /
count) / sum_value_in_day * 100
RMS_daily_error[:, index, 1] = RMS_daily_error[:, index, 0]
BIAS_daily_error[:, index, 1] = BIAS_daily_error[:, index, 0]
print 'RMS :', RMS_daily_error[:, index, 0]
print 'BIAS', BIAS_daily_error[:, index, 0]
print 'sum value in day: ', sum_value_in_day[:]
show("\rDiary RMS daily error: %.2f\n" %
(RMS_daily_error[:, index, 0]).mean())
nc.sync(root)
nc.close(root)
import glob
from osgeo import gdal
from libs.file import netcdf as nc
files = glob.glob("libs/linke/tifs/*.tif")
shape = gdal.Open(files[0]).ReadAsArray().shape
root, n = nc.open("linketurbidity.nc")
nc.getdim(root, "northing", shape[0])
nc.getdim(root, "easting", shape[1])
nc.getdim(root, "monthing")
linke = nc.getvar(root, "linketurbidity", "f4", ("monthing", "northing", "easting",), 4)
for i in range(len(files)):
print i, '->', files[i]
linke[i,:] = gdal.Open(files[i]).ReadAsArray()
nc.close(root)