def test_issue_189(self):
fn = '/mnt/10.11.12.232/sat_downloads_asar/level-0/2010-01/descending/VV/gsar_rvl/RVL_ASA_WS_20100110211812087.gsar'
if doppler_installed:
n = Doppler(fn)
xlon, xlat = n.get_corners()
d = Domain(
NSR(3857), '-lle %f %f %f %f -tr 1000 1000' %
(xlon.min(), xlat.min(), xlon.max(), xlat.max()))
n.reproject(d, eResampleAlg=1, tps=True)
inci = n['incidence_angle']
def test_issue_189(self):
fn = '/mnt/10.11.12.232/sat_downloads_asar/level-0/2010-01/descending/VV/gsar_rvl/RVL_ASA_WS_20100110211812087.gsar'
if doppler_installed:
n = Doppler(fn)
xlon, xlat = n.get_corners()
d = Domain(NSR(3857),
'-lle %f %f %f %f -tr 1000 1000' % (
xlon.min(), xlat.min(), xlon.max(), xlat.max()))
n.reproject(d, eResampleAlg=1, tps=True)
inci = n['incidence_angle']
def process(self, uri, *args, **kwargs):
""" Create data products
"""
ds, created = self.get_or_create(uri, *args, **kwargs)
fn = nansat_filename(uri)
swath_data = {}
# Read subswaths
for i in range(self.N_SUBSWATHS):
swath_data[i] = Doppler(fn, subswath=i)
# Get module name
mm = self.__module__.split('.')
module = '%s.%s' % (mm[0], mm[1])
# Set media path (where images will be stored)
mp = media_path(module, swath_data[i].filename)
# Set product path (where netcdf products will be stored)
ppath = product_path(module, swath_data[i].filename)
# Loop subswaths, process each of them and create figures for display with leaflet
for i in range(self.N_SUBSWATHS):
is_corrupted = False
# Check if the file is corrupted
try:
inci = swath_data[i]['incidence_angle']
# TODO: What kind of exception ?
except:
is_corrupted = True
continue
# Add Doppler anomaly
swath_data[i].add_band(array=swath_data[i].anomaly(), parameters={
'wkv':
'anomaly_of_surface_backwards_doppler_centroid_frequency_shift_of_radar_wave'
})
# Get band number of DC freq, then DC polarisation
band_number = swath_data[i]._get_band_number({
'standard_name': 'surface_backwards_doppler_centroid_frequency_shift_of_radar_wave',
})
pol = swath_data[i].get_metadata(bandID=band_number, key='polarization')
# Calculate total geophysical Doppler shift
fdg = swath_data[i].geophysical_doppler_shift()
swath_data[i].add_band(
array=fdg,
parameters={
'wkv': 'surface_backwards_doppler_frequency_shift_of_radar_wave_due_to_surface_velocity'
})
# Set filename of exported netcdf
fn = os.path.join(ppath,
os.path.basename(swath_data[i].filename).split('.')[0]
+ 'subswath%d.nc' % i)
# Set filename of original gsar file in metadata
swath_data[i].set_metadata(key='Originating file',
value=swath_data[i].filename)
# Export data to netcdf
print('Exporting %s (subswath %d)' % (swath_data[i].filename, i))
swath_data[i].export(filename=fn)
# Add netcdf uri to DatasetURIs
ncuri = os.path.join('file://localhost', fn)
new_uri, created = DatasetURI.objects.get_or_create(uri=ncuri,
dataset=ds)
# Reproject to leaflet projection
xlon, xlat = swath_data[i].get_corners()
d = Domain(NSR(3857),
'-lle %f %f %f %f -tr 1000 1000'
% (xlon.min(), xlat.min(), xlon.max(), xlat.max()))
swath_data[i].reproject(d, eResampleAlg=1, tps=True)
# Check if the reprojection failed
try:
inci = swath_data[i]['incidence_angle']
except:
is_corrupted = True
warnings.warn('Could not read incidence angles - reprojection failed')
continue
# Create visualizations of the following bands (short_names)
ingest_creates = ['valid_doppler',
'valid_land_doppler',
'valid_sea_doppler',
'dca',
'fdg']
for band in ingest_creates:
filename = '%s_subswath_%d.png' % (band, i)
# check uniqueness of parameter
param = Parameter.objects.get(short_name=band)
fig = swath_data[i].write_figure(
os.path.join(mp, filename),
bands=band,
mask_array=swath_data[i]['swathmask'],
mask_lut={0: [128, 128, 128]},
transparency=[128, 128, 128])
if type(fig) == Figure:
print 'Created figure of subswath %d, band %s' % (i, band)
else:
warnings.warn('Figure NOT CREATED')
# Get or create DatasetParameter
dsp, created = DatasetParameter.objects.get_or_create(dataset=ds,
parameter=param)
# Create GeographicLocation for the visualization object
geom, created = GeographicLocation.objects.get_or_create(
geometry=WKTReader().read(swath_data[i].get_border_wkt()))
# Create Visualization
vv, created = Visualization.objects.get_or_create(
uri='file://localhost%s/%s' % (mp, filename),
title='%s (swath %d)' % (param.standard_name, i + 1),
geographic_location=geom
)
# Create VisualizationParameter
vp, created = VisualizationParameter.objects.get_or_create(
visualization=vv,
ds_parameter=dsp
)
# TODO: consider merged figures like Jeong-Won has added in the development branch
return ds, not is_corrupted
def create_merged_swaths(self, ds, EPSG=4326, **kwargs):
"""Merge swaths, add dataseturi, and return Nansat object.
EPSG options:
- 4326: WGS 84 / longlat
- 3995: WGS 84 / Arctic Polar Stereographic
"""
nn = {}
nn[0] = Doppler(
nansat_filename(
ds.dataseturi_set.get(uri__endswith='%d.nc' % 0).uri))
lon0, lat0 = nn[0].get_geolocation_grids()
nn[1] = Doppler(
nansat_filename(
ds.dataseturi_set.get(uri__endswith='%d.nc' % 1).uri))
lon1, lat1 = nn[1].get_geolocation_grids()
nn[2] = Doppler(
nansat_filename(
ds.dataseturi_set.get(uri__endswith='%d.nc' % 2).uri))
lon2, lat2 = nn[2].get_geolocation_grids()
nn[3] = Doppler(
nansat_filename(
ds.dataseturi_set.get(uri__endswith='%d.nc' % 3).uri))
lon3, lat3 = nn[3].get_geolocation_grids()
nn[4] = Doppler(
nansat_filename(
ds.dataseturi_set.get(uri__endswith='%d.nc' % 4).uri))
lon4, lat4 = nn[4].get_geolocation_grids()
connection.close()
dlon = np.mean([
np.abs(np.mean(np.gradient(lon0, axis=1))),
np.abs(np.mean(np.gradient(lon1, axis=1))),
np.abs(np.mean(np.gradient(lon2, axis=1))),
np.abs(np.mean(np.gradient(lon3, axis=1))),
np.abs(np.mean(np.gradient(lon4, axis=1)))
])
nx = len(
np.arange(
np.array([
lon0.min(),
lon1.min(),
lon2.min(),
lon3.min(),
lon4.min()
]).min(),
np.array([
lon0.max(),
lon1.max(),
lon2.max(),
lon3.max(),
lon4.max()
]).max(), dlon))
dlat = np.mean([
np.abs(np.mean(np.gradient(lat0, axis=0))),
np.abs(np.mean(np.gradient(lat1, axis=0))),
np.abs(np.mean(np.gradient(lat2, axis=0))),
np.abs(np.mean(np.gradient(lat3, axis=0))),
np.abs(np.mean(np.gradient(lat4, axis=0)))
])
ny = len(
np.arange(
np.array([
lat0.min(),
lat1.min(),
lat2.min(),
lat3.min(),
lat4.min()
]).min(),
np.array([
lat0.max(),
lat1.max(),
lat2.max(),
lat3.max(),
lat4.max()
]).max(), dlat))
if ny is None:
ny = np.array([
nn[0].shape()[0], nn[1].shape()[0], nn[2].shape()[0],
nn[3].shape()[0], nn[4].shape()[0]
]).max()
## DETTE VIRKER IKKE..
#sensor_view = np.sort(
# np.append(np.append(np.append(np.append(
# nn[0]['sensor_view'][0,:],
# nn[1]['sensor_view'][0,:]),
# nn[2]['sensor_view'][0,:]),
# nn[3]['sensor_view'][0,:]),
# nn[4]['sensor_view'][0,:]))
#nx = sensor_view.size
#x = np.arange(nx)
#def func(x, a, b, c, d):
# return a*x**3+b*x**2+c*x+d
#def linear_func(x, a, b):
# return a*x + b
#azimuth_time = np.sort(
# np.append(np.append(np.append(np.append(
# nn[0].get_azimuth_time(),
# nn[1].get_azimuth_time()),
# nn[2].get_azimuth_time()),
# nn[3].get_azimuth_time()),
# nn[4].get_azimuth_time()))
#dt = azimuth_time.max() - azimuth_time[0]
#tt = np.arange(0, dt, dt/ny)
#tt = np.append(np.array([-dt/ny], dtype='<m8[us]'), tt)
#tt = np.append(tt, tt[-1]+np.array([dt/ny, 2*dt/ny], dtype='<m8[us]'))
#ny = len(tt)
## AZIMUTH_TIME
#azimuth_time = (np.datetime64(azimuth_time[0])+tt).astype(datetime)
#popt, pcov = curve_fit(func, x, sensor_view)
## SENSOR VIEW ANGLE
#alpha = np.ones((ny, sensor_view.size))*np.deg2rad(func(x, *popt))
#range_time = np.sort(
# np.append(np.append(np.append(np.append(
# nn[0].get_range_time(),
# nn[1].get_range_time()),
# nn[2].get_range_time()),
# nn[3].get_range_time()),
# nn[4].get_range_time()))
#popt, pcov = curve_fit(linear_func, x, range_time)
## RANGE_TIME
#range_time = linear_func(x, *popt)
#ecefPos, ecefVel = Doppler.orbital_state_vectors(azimuth_time)
#eciPos, eciVel = ecef2eci(ecefPos, ecefVel, azimuth_time)
## Get satellite hour angle
#satHourAng = np.deg2rad(Doppler.satellite_hour_angle(azimuth_time, ecefPos, ecefVel))
## Get attitude from the Envisat yaw steering law
#psi, gamma, phi = np.deg2rad(Doppler.orbital_attitude_vectors(azimuth_time, satHourAng))
#U1, AX1, S1 = Doppler.step_one_calculations(alpha, psi, gamma, phi, eciPos)
#S2, U2, AX2 = Doppler.step_two_calculations(satHourAng, S1, U1, AX1)
#S3, U3, AX3 = Doppler.step_three_a_calculations(eciPos, eciVel, S2, U2, AX2)
#U3g = Doppler.step_three_b_calculations(S3, U3, AX3)
#P3, U3g, lookAng = Doppler.step_four_calculations(S3, U3g, AX3, range_time)
#dcm = dcmeci2ecef(azimuth_time, 'IAU-2000/2006')
#lat = np.zeros((ny, nx))
#lon = np.zeros((ny, nx))
#alt = np.zeros((ny, nx))
#for i in range(P3.shape[1]):
# ecefPos = np.matmul(dcm[0], P3[:,i,:,0, np.newaxis])
# lla = ecef2lla(ecefPos)
# lat[:,i] = lla[:,0]
# lon[:,i] = lla[:,1]
# alt[:,i] = lla[:,2]
#lon = lon.round(decimals=5)
#lat = lat.round(decimals=5)
# DETTE VIRKER:
lonmin = np.array(
[lon0.min(),
lon1.min(),
lon2.min(),
lon3.min(),
lon4.min()]).min()
lonmax = np.array(
[lon0.max(),
lon1.max(),
lon2.max(),
lon3.max(),
lon4.max()]).max()
latmin = np.array(
[lat0.min(),
lat1.min(),
lat2.min(),
lat3.min(),
lat4.min()]).min()
latmax = np.array(
[lat0.max(),
lat1.max(),
lat2.max(),
lat3.max(),
lat4.max()]).max()
if nx is None:
nx = nn[0].shape()[1] + nn[1].shape()[1] + nn[2].shape()[1] + nn[3].shape()[1] + \
nn[4].shape()[1]
# prepare geospatial grid
merged = Nansat.from_domain(
Domain(
NSR(EPSG), '-lle %f %f %f %f -ts %d %d' %
(lonmin, latmin, lonmax, latmax, nx, ny)))
## DETTE VIRKER IKKE..
#merged = Nansat.from_domain(Domain.from_lonlat(lon, lat, add_gcps=False))
#merged.add_band(array = np.rad2deg(alpha), parameters={'wkv': 'sensor_view'})
dfdg = np.ones((self.N_SUBSWATHS)) * 5 # Hz (5 Hz a priori)
for i in range(self.N_SUBSWATHS):
dfdg[i] = nn[i].get_uncertainty_of_fdg()
nn[i].reproject(merged, tps=True, resample_alg=1, block_size=2)
# Initialize band arrays
inc = np.ones(
(self.N_SUBSWATHS, merged.shape()[0], merged.shape()[1])) * np.nan
fdg = np.ones(
(self.N_SUBSWATHS, merged.shape()[0], merged.shape()[1])) * np.nan
ur = np.ones(
(self.N_SUBSWATHS, merged.shape()[0], merged.shape()[1])) * np.nan
valid_sea_dop = np.ones(
(self.N_SUBSWATHS, merged.shape()[0], merged.shape()[1])) * np.nan
std_fdg = np.ones(
(self.N_SUBSWATHS, merged.shape()[0], merged.shape()[1])) * np.nan
std_ur = np.ones(
(self.N_SUBSWATHS, merged.shape()[0], merged.shape()[1])) * np.nan
for ii in range(self.N_SUBSWATHS):
inc[ii] = nn[ii]['incidence_angle']
fdg[ii] = nn[ii]['fdg']
ur[ii] = nn[ii]['Ur']
valid_sea_dop[ii] = nn[ii]['valid_sea_doppler']
# uncertainty of fdg is a scalar
std_fdg[ii][valid_sea_dop[ii] == 1] = dfdg[ii]
# uncertainty of ur
std_ur[ii] = nn[ii].get_uncertainty_of_radial_current(dfdg[ii])
# Calculate incidence angle as a simple average
mean_inc = np.nanmean(inc, axis=0)
merged.add_band(array=mean_inc,
parameters={
'name': 'incidence_angle',
'wkv': 'angle_of_incidence'
})
# Calculate fdg as weighted average
mean_fdg = nansumwrapper((fdg/np.square(std_fdg)).data, axis=0) / \
nansumwrapper((1./np.square(std_fdg)).data, axis=0)
merged.add_band(
array=mean_fdg,
parameters={
'name':
'fdg',
'wkv':
'surface_backwards_doppler_frequency_shift_of_radar_wave_due_to_surface_velocity'
})
# Standard deviation of fdg
std_mean_fdg = np.sqrt(1. / nansumwrapper(
(1. / np.square(std_fdg)).data, axis=0))
merged.add_band(array=std_mean_fdg, parameters={'name': 'std_fdg'})
# Calculate ur as weighted average
mean_ur = nansumwrapper((ur/np.square(std_ur)).data, axis=0) / \
nansumwrapper((1./np.square(std_ur)).data, axis=0)
merged.add_band(array=mean_ur, parameters={
'name': 'Ur',
})
# Standard deviation of Ur
std_mean_ur = np.sqrt(1. / nansumwrapper(
(1. / np.square(std_ur)).data, axis=0))
merged.add_band(array=std_mean_ur, parameters={'name': 'std_ur'})
# Band of valid pixels
vsd = np.nanmin(valid_sea_dop, axis=0)
merged.add_band(array=vsd, parameters={
'name': 'valid_sea_doppler',
})
# Add file to db
fn = os.path.join(
product_path(
self.module_name(),
nansat_filename(
ds.dataseturi_set.get(uri__endswith='.gsar').uri)),
os.path.basename(
nansat_filename(
ds.dataseturi_set.get(
uri__endswith='.gsar').uri)).split('.')[0] +
'_merged.nc')
merged.export(filename=fn)
ncuri = 'file://localhost' + fn
new_uri, created = DatasetURI.objects.get_or_create(uri=ncuri,
dataset=ds)
connection.close()
return merged
def process(self, ds, force=False, *args, **kwargs):
""" Create data products
Returns
=======
ds : geospaas.catalog.models.Dataset
processed : Boolean
Flag to indicate if the dataset was processed or not
"""
swath_data = {}
# Set media path (where images will be stored)
mp = media_path(
self.module_name(),
nansat_filename(ds.dataseturi_set.get(uri__endswith='.gsar').uri))
# Set product path (where netcdf products will be stored)
ppath = product_path(
self.module_name(),
nansat_filename(ds.dataseturi_set.get(uri__endswith='.gsar').uri))
# Read subswaths
dss = {1: None, 2: None, 3: None, 4: None, 5: None}
processed = [True, True, True, True, True]
failing = [False, False, False, False, False]
for i in range(self.N_SUBSWATHS):
# Check if the data has already been processed
try:
fn = nansat_filename(
ds.dataseturi_set.get(uri__endswith='%d.nc' % i).uri)
except DatasetURI.DoesNotExist:
processed[i] = False
else:
dd = Nansat(fn)
try:
std_Ur = dd['std_Ur']
except ValueError:
processed[i] = False
if processed[i] and not force:
continue
# Process from scratch to avoid duplication of bands
fn = nansat_filename(
ds.dataseturi_set.get(uri__endswith='.gsar').uri)
try:
dd = Doppler(fn, subswath=i)
except Exception as e:
logging.error('%s (Filename, subswath [1-5]): (%s, %d)' %
(str(e), fn, i + 1))
failing[i] = True
continue
# Check if the file is corrupted
try:
inc = dd['incidence_angle']
except Exception as e:
logging.error('%s (Filename, subswath [1-5]): (%s, %d)' %
(str(e), fn, i + 1))
failing[i] = True
continue
dss[i + 1] = dd
if all(processed) and not force:
logging.info("%s: The dataset has already been processed." %
nansat_filename(
ds.dataseturi_set.get(uri__endswith='.gsar').uri))
return ds, False
if all(failing):
logging.error(
"Processing of all subswaths is failing: %s" % nansat_filename(
ds.dataseturi_set.get(uri__endswith='.gsar').uri))
return ds, False
if any(failing):
logging.error(
"Some but not all subswaths processed: %s" % nansat_filename(
ds.dataseturi_set.get(uri__endswith='.gsar').uri))
return ds, False
logging.info(
"Processing %s" %
nansat_filename(ds.dataseturi_set.get(uri__endswith='.gsar').uri))
# Loop subswaths, process each of them
processed = False
def get_overlap(d1, d2):
b1 = d1.get_border_geometry()
b2 = d2.get_border_geometry()
intersection = b1.Intersection(b2)
lo1, la1 = d1.get_geolocation_grids()
overlap = np.zeros(lo1.shape)
for i in range(lo1.shape[0]):
for j in range(lo1.shape[1]):
wkt_point = 'POINT(%.5f %.5f)' % (lo1[i, j], la1[i, j])
overlap[i, j] = intersection.Contains(
ogr.CreateGeometryFromWkt(wkt_point))
return overlap
for uri in ds.dataseturi_set.filter(uri__endswith='.nc'):
logging.debug("%s" % nansat_filename(uri.uri))
# Find pixels in dss[1] which overlap with pixels in dss[2]
overlap12 = get_overlap(dss[1], dss[2])
# Find pixels in dss[2] which overlap with pixels in dss[1]
overlap21 = get_overlap(dss[2], dss[1])
# and so on..
overlap23 = get_overlap(dss[2], dss[3])
overlap32 = get_overlap(dss[3], dss[2])
overlap34 = get_overlap(dss[3], dss[4])
overlap43 = get_overlap(dss[4], dss[3])
overlap45 = get_overlap(dss[4], dss[5])
overlap54 = get_overlap(dss[5], dss[4])
# Get range bias corrected Doppler
fdg = {}
fdg[1] = dss[1].anomaly() - dss[1].range_bias()
fdg[2] = dss[2].anomaly() - dss[2].range_bias()
fdg[3] = dss[3].anomaly() - dss[3].range_bias()
fdg[4] = dss[4].anomaly() - dss[4].range_bias()
fdg[5] = dss[5].anomaly() - dss[5].range_bias()
# Get median values at overlapping borders
median12 = np.nanmedian(fdg[1][np.where(overlap12)])
median21 = np.nanmedian(fdg[2][np.where(overlap21)])
median23 = np.nanmedian(fdg[2][np.where(overlap23)])
median32 = np.nanmedian(fdg[3][np.where(overlap32)])
median34 = np.nanmedian(fdg[3][np.where(overlap34)])
median43 = np.nanmedian(fdg[4][np.where(overlap43)])
median45 = np.nanmedian(fdg[4][np.where(overlap45)])
median54 = np.nanmedian(fdg[5][np.where(overlap54)])
# Adjust levels to align at subswath borders
fdg[1] -= median12 - np.nanmedian(np.array([median12, median21]))
fdg[2] -= median21 - np.nanmedian(np.array([median12, median21]))
fdg[1] -= median23 - np.nanmedian(np.array([median23, median32]))
fdg[2] -= median23 - np.nanmedian(np.array([median23, median32]))
fdg[3] -= median32 - np.nanmedian(np.array([median23, median32]))
fdg[1] -= median34 - np.nanmedian(np.array([median34, median43]))
fdg[2] -= median34 - np.nanmedian(np.array([median34, median43]))
fdg[3] -= median34 - np.nanmedian(np.array([median34, median43]))
fdg[4] -= median43 - np.nanmedian(np.array([median34, median43]))
fdg[1] -= median45 - np.nanmedian(np.array([median45, median54]))
fdg[2] -= median45 - np.nanmedian(np.array([median45, median54]))
fdg[3] -= median45 - np.nanmedian(np.array([median45, median54]))
fdg[4] -= median45 - np.nanmedian(np.array([median45, median54]))
fdg[5] -= median54 - np.nanmedian(np.array([median45, median54]))
# Correct by land or mean fww
try:
wind_fn = nansat_filename(
Dataset.objects.get(
source__platform__short_name='ERA15DAS',
time_coverage_start__lte=ds.time_coverage_end,
time_coverage_end__gte=ds.time_coverage_start).
dataseturi_set.get().uri)
except Exception as e:
logging.error(
"%s - in search for ERA15DAS data (%s, %s, %s) " %
(str(e),
nansat_filename(
ds.dataseturi_set.get(uri__endswith=".gsar").uri),
ds.time_coverage_start, ds.time_coverage_end))
return ds, False
connection.close()
land = np.array([])
fww = np.array([])
offset_corrected = 0
for key in dss.keys():
land = np.append(
land, fdg[key][dss[key]['valid_land_doppler'] == 1].flatten())
if land.any():
logging.info('Using land for bias corrections')
land_bias = np.nanmedian(land)
offset_corrected = 1
else:
logging.info('Using CDOP wind-waves Doppler for bias corrections')
# correct by mean wind doppler
for key in dss.keys():
ff = fdg[key].copy()
# do CDOP correction
ff[ dss[key]['valid_sea_doppler']==1 ] = \
ff[ dss[key]['valid_sea_doppler']==1 ] \
- dss[key].wind_waves_doppler(wind_fn)[0][ dss[key]['valid_sea_doppler']==1 ]
ff[dss[key]['valid_doppler'] == 0] = np.nan
fww = np.append(fww, ff.flatten())
land_bias = np.nanmedian(fww)
if np.isnan(land_bias):
offset_corrected = 0
raise Exception('land bias is NaN...')
else:
offset_corrected = 1
for key in dss.keys():
fdg[key] -= land_bias
# Set unrealistically high/low values to NaN (ref issue #4 and #5)
fdg[key][fdg[key] < -100] = np.nan
fdg[key][fdg[key] > 100] = np.nan
# Add fdg[key] as band
dss[key].add_band(
array=fdg[key],
parameters={
'wkv':
'surface_backwards_doppler_frequency_shift_of_radar_wave_due_to_surface_velocity',
'offset_corrected': str(offset_corrected)
})
# Add Doppler anomaly
dss[key].add_band(
array=dss[key].anomaly(),
parameters={
'wkv':
'anomaly_of_surface_backwards_doppler_centroid_frequency_shift_of_radar_wave'
})
# Add wind doppler and its uncertainty as bands
fww, dfww = dss[key].wind_waves_doppler(wind_fn)
dss[key].add_band(
array=fww,
parameters={
'wkv':
'surface_backwards_doppler_frequency_shift_of_radar_wave_due_to_wind_waves'
})
dss[key].add_band(array=dfww, parameters={'name': 'std_fww'})
# Calculate range current velocity component
v_current, std_v, offset_corrected = \
dss[key].surface_radial_doppler_sea_water_velocity(wind_fn, fdg=fdg[key])
dss[key].add_band(array=v_current,
parameters={
'wkv':
'surface_radial_doppler_sea_water_velocity',
'offset_corrected': str(offset_corrected)
})
dss[key].add_band(array=std_v, parameters={'name': 'std_Ur'})
# Set satellite pass
lon, lat = dss[key].get_geolocation_grids()
gg = np.gradient(lat, axis=0)
dss[key].add_band(array=gg,
parameters={
'name':
'sat_pass',
'comment':
'ascending pass is >0, descending pass is <0'
})
history_message = (
'sar_doppler.models.Dataset.objects.process("%s") '
'[geospaas sar_doppler version %s]' %
(ds, os.getenv('GEOSPAAS_SAR_DOPPLER_VERSION', 'dev')))
new_uri, created = self.export2netcdf(
dss[key], ds, history_message=history_message)
processed = True
m = self.create_merged_swaths(ds)
return ds, processed
def calc_mean_doppler(datetime_start=timezone.datetime(2010,1,1,
tzinfo=timezone.utc), datetime_end=timezone.datetime(2010,2,1,
tzinfo=timezone.utc), domain=Domain(NSR().wkt,
'-te 10 -44 40 -30 -tr 0.05 0.05')):
geometry = WKTReader().read(domain.get_border_wkt(nPoints=1000))
ds = Dataset.objects.filter(entry_title__contains='Doppler',
time_coverage_start__range=[datetime_start, datetime_end],
geographic_location__geometry__intersects=geometry)
Va = np.zeros(domain.shape())
Vd = np.zeros(domain.shape())
ca = np.zeros(domain.shape())
cd = np.zeros(domain.shape())
sa = np.zeros(domain.shape())
sd = np.zeros(domain.shape())
sum_var_inv_a = np.zeros(domain.shape())
sum_var_inv_d = np.zeros(domain.shape())
for dd in ds:
uris = dd.dataseturi_set.filter(uri__endswith='nc')
for uri in uris:
dop = Doppler(uri.uri)
# Consider skipping swath 1 and possibly 2...
dop.reproject(domain)
# TODO: HARDCODING - MUST BE IMPROVED
satpass = dop.get_metadata(key='Originating file').split('/')[6]
if satpass=='ascending':
try:
v_ai = dop['Ur']
v_ai[np.abs(v_ai)>3] = np.nan
except:
# subswath doesn't cover the given domain
continue
# uncertainty:
# 5 Hz - TODO: estimate this correctly...
sigma_ai = -np.pi*np.ones(dop.shape())*5./(112*np.sin(dop['incidence_angle']*np.pi/180.))
alpha_i = -dop['sensor_azimuth']*np.pi/180.
Va = np.nansum(np.append(np.expand_dims(Va, 2),
np.expand_dims(v_ai/np.square(sigma_ai), 2), axis=2),
axis=2)
ca = np.nansum(np.append(np.expand_dims(ca, 2),
np.expand_dims(np.cos(alpha_i)/np.square(sigma_ai), 2),
axis=2), axis=2)
sa = np.nansum(np.append(np.expand_dims(sa, 2),
np.expand_dims(np.sin(alpha_i)/np.square(sigma_ai), 2),
axis=2), axis=2)
sum_var_inv_a =np.nansum(np.append(np.expand_dims(sum_var_inv_a, 2),
np.expand_dims(1./np.square(sigma_ai), 2), axis=2),
axis=2)
else:
try:
v_dj = -dop['Ur']
v_dj[np.abs(v_dj)>3] = np.nan
except:
# subswath doesn't cover the given domain
continue
# 5 Hz - TODO: estimate this correctly...
sigma_dj = -np.pi*np.ones(dop.shape())*5./(112*np.sin(dop['incidence_angle']*np.pi/180.))
delta_j = (dop['sensor_azimuth']-180.)*np.pi/180.
Vd = np.nansum(np.append(np.expand_dims(Vd, 2),
np.expand_dims(v_dj/np.square(sigma_dj), 2), axis=2),
axis=2)
cd = np.nansum(np.append(np.expand_dims(cd, 2),
np.expand_dims(np.cos(delta_j)/np.square(sigma_dj), 2),
axis=2), axis=2)
sd = np.nansum(np.append(np.expand_dims(sd, 2),
np.expand_dims(np.sin(delta_j)/np.square(sigma_dj), 2),
axis=2), axis=2)
sum_var_inv_d = np.nansum(np.append(
np.expand_dims(sum_var_inv_d, 2), np.expand_dims(
1./np.square(sigma_ai), 2), axis=2), axis=2)
u = (Va*sd + Vd*sa)/(sa*cd + sd*ca)
v = (Va*cd - Vd*ca)/(sa*cd + sd*ca)
sigma_u = np.sqrt(np.square(sd)*sum_var_inv_a +
np.square(sa)*sum_var_inv_d) / (sa*cd + sd*ca)
sigma_v = np.sqrt(np.square(cd)*sum_var_inv_a +
np.square(ca)*sum_var_inv_d) / (sa*cd + sd*ca)
nu = Nansat(array=u, domain=domain)
nmap=Nansatmap(nu, resolution='h')
nmap.pcolormesh(nu[1], vmin=-1.5, vmax=1.5, cmap='bwr')
nmap.add_colorbar()
nmap.draw_continents()
nmap.fig.savefig('/vagrant/shared/unwasc.png', bbox_inches='tight')
def update_geophysical_doppler(dopplerFile, t0, t1, swath, sensor='ASAR',
platform='ENVISAT'):
dop2correct = Doppler(dopplerFile)
bandnum = dop2correct._get_band_number({
'standard_name':
'surface_backwards_doppler_centroid_frequency_shift_of_radar_wave'
})
polarization = dop2correct.get_metadata(bandID=bandnum, key='polarization')
lon,lat = dop2correct.get_geolocation_grids()
indmidaz = lat.shape[0]/2
indmidra = lat.shape[1]/2
if lat[indmidaz,indmidra]>lat[0,indmidra]:
use_pass = '******'
else:
use_pass = '******'
# Get datasets
DS = Dataset.objects.filter(source__platform__short_name=platform,
source__instrument__short_name=sensor)
dopDS = DS.filter(
parameters__short_name = 'dca',
time_coverage_start__gte = t0,
time_coverage_start__lt = t1
)
swath_files = []
for dd in dopDS:
try:
fn = dd.dataseturi_set.get(
uri__endswith='subswath%s.nc' %swath).uri
except DatasetURI.DoesNotExist:
continue
n = Doppler(fn)
try:
dca = n.anomaly(pol=polarization)
except OptionError: # wrong polarization..
continue
lon,lat=n.get_geolocation_grids()
indmidaz = lat.shape[0]/2
indmidra = lat.shape[1]/2
if lat[indmidaz,indmidra]>lat[0,indmidra]:
orbit_pass = '******'
else:
orbit_pass = '******'
if use_pass==orbit_pass:
swath_files.append(fn)
valid_land = np.array([])
valid = np.array([])
for ff in swath_files:
n = Nansat(ff)
view_bandnum = n._get_band_number({
'standard_name': 'sensor_view_angle'
})
std_bandnum = n._get_band_number({
'standard_name': \
'standard_deviation_of_surface_backwards_doppler_centroid_frequency_shift_of_radar_wave',
})
pol = n.get_metadata(bandID=std_bandnum, key='polarization')
# For checking when antenna pattern changes
if valid.shape==(0,):
valid = n['valid_doppler']
dca0 = n['dca']
dca0[n['valid_doppler']==0] = np.nan
dca0[n['valid_sea_doppler']==1] = dca0[n['valid_sea_doppler']==1] - \
n['fww'][n['valid_sea_doppler']==1]
view_angle0 = n[view_bandnum]
else:
validn = n['valid_doppler']
dca0n = n['dca']
dca0n[n['valid_doppler']==0] = np.nan
dca0n[n['valid_sea_doppler']==1] = dca0n[n['valid_sea_doppler']==1] - \
n['fww'][n['valid_sea_doppler']==1]
view_angle0n = n[view_bandnum]
if not validn.shape==valid.shape:
if validn.shape[1] > valid.shape[1]:
valid = np.resize(valid, (valid.shape[0], validn.shape[1]))
dca0 = np.resize(dca0, (dca0.shape[0], dca0n.shape[1]))
view_angle0 = np.resize(view_angle0,
(view_angle0.shape[0], view_angle0n.shape[1]))
else:
validn = np.resize(validn, (validn.shape[0],
valid.shape[1]))
dca0n = np.resize(dca0n, (dca0n.shape[0], dca0.shape[1]))
view_angle0n = np.resize(view_angle0n,
(view_angle0n.shape[0], view_angle0.shape[1]))
valid = np.concatenate((valid, validn))
dca0 = np.concatenate((dca0, dca0n))
view_angle0 = np.concatenate((view_angle0, view_angle0n))
if valid_land.shape==(0,):
valid_land = n['valid_land_doppler'][n['valid_land_doppler'].any(axis=1)]
dca = n['dca'][n['valid_land_doppler'].any(axis=1)]
view_angle = n[view_bandnum][n['valid_land_doppler'].any(axis=1)]
std_dca = n[std_bandnum][n['valid_land_doppler'].any(axis=1)]
else:
vn = n['valid_land_doppler'][n['valid_land_doppler'].any(axis=1)]
dcan = n['dca'][n['valid_land_doppler'].any(axis=1)]
view_angle_n = n[view_bandnum][n['valid_land_doppler'].any(axis=1)]
std_dca_n = n[std_bandnum][n['valid_land_doppler'].any(axis=1)]
if not vn.shape==valid_land.shape:
# Resize arrays - just for visual inspection. Actual interpolation
# is view angle vs doppler anomaly
if vn.shape[1] > valid_land.shape[1]:
valid_land = np.resize(valid_land, (valid_land.shape[0],
vn.shape[1]))
dca = np.resize(dca, (dca.shape[0],
vn.shape[1]))
view_angle = np.resize(view_angle, (view_angle.shape[0],
vn.shape[1]))
std_dca = np.resize(std_dca, (std_dca.shape[0],
vn.shape[1]))
if vn.shape[1] < valid_land.shape[1]:
vn = np.resize(vn, (vn.shape[0], valid_land.shape[1]))
dcan = np.resize(dcan, (dcan.shape[0], valid_land.shape[1]))
view_angle_n = np.resize(view_angle_n, (view_angle_n.shape[0], valid_land.shape[1]))
std_dca_n = np.resize(std_dca_n, (std_dca_n.shape[0], valid_land.shape[1]))
valid_land = np.concatenate((valid_land, vn))
dca = np.concatenate((dca, dcan))
view_angle = np.concatenate((view_angle, view_angle_n))
std_dca = np.concatenate((std_dca, std_dca_n))
view_angle0 = view_angle0.flatten()
dca0 = dca0.flatten()
view_angle0 = np.delete(view_angle0, np.where(np.isnan(dca0)))
dca0 = np.delete(dca0, np.where(np.isnan(dca0)))
ind = np.argsort(view_angle0)
view_angle0 = view_angle0[ind]
dca0 = dca0[ind]
# Set dca, view_angle and std_dca to nan where not land
dca[valid_land==0] = np.nan
std_dca[valid_land==0] = np.nan
view_angle[valid_land==0] = np.nan
dca = dca.flatten()
std_dca = std_dca.flatten()
view_angle = view_angle.flatten()
dca = np.delete(dca, np.where(np.isnan(dca)))
std_dca = np.delete(std_dca, np.where(np.isnan(std_dca)))
view_angle = np.delete(view_angle, np.where(np.isnan(view_angle)))
ind = np.argsort(view_angle)
view_angle = view_angle[ind]
dca = dca[ind]
std_dca = std_dca[ind]
freqLims = [-200,200]
# Show this in presentation:
plt.subplot(2,1,1)
count, anglebins, dcabins, im = plt.hist2d(view_angle0, dca0, 100, cmin=1,
range=[[np.min(view_angle), np.max(view_angle)], freqLims])
plt.colorbar()
plt.title('Wind Doppler subtracted')
plt.subplot(2,1,2)
count, anglebins, dcabins, im = plt.hist2d(view_angle, dca, 100, cmin=1,
range=[[np.min(view_angle), np.max(view_angle)], freqLims])
plt.colorbar()
plt.title('Doppler over land')
#plt.show()
plt.close()
countLims = 200
#{
# 0: 600,
# 1: 250,
# 2: 500,
# 3: 140,
# 4: 130,
#}
dcabins_grid, anglebins_grid = np.meshgrid(dcabins[:-1], anglebins[:-1])
anglebins_vec = anglebins_grid[count>countLims]
dcabins_vec = dcabins_grid[count>countLims]
#anglebins_vec = anglebins_grid[count>countLims[swath]]
#dcabins_vec = dcabins_grid[count>countLims[swath]]
va4interp = []
rb4interp = []
std_rb4interp = []
for i in range(len(anglebins)-1):
if i==0:
ind0 = 0
else:
ind0 = np.where(view_angle>anglebins[i])[0][0]
ind1 = np.where(view_angle<=anglebins[i+1])[0][-1]
va4interp.append(np.mean(view_angle[ind0:ind1]))
rb4interp.append(np.median(dca[ind0:ind1]))
std_rb4interp.append(np.std(dca[ind0:ind1]))
va4interp = np.array(va4interp)
rb4interp = np.array(rb4interp)
std_rb4interp = np.array(std_rb4interp)
van = dop2correct['sensor_view']
rbfull = van.copy()
rbfull[:,:] = np.nan
# Is there a more efficient method than looping?
import time
start_time = time.time()
for ii in range(len(anglebins)-1):
vaii0 = anglebins[ii]
vaii1 = anglebins[ii+1]
rbfull[(van>=vaii0) & (van<=vaii1)] = \
np.median(dca[(view_angle>=vaii0) & (view_angle<=vaii1)])
#print("--- %s seconds ---" % (time.time() - start_time))
plt.plot(np.mean(van, axis=0), np.mean(rbfull, axis=0), '.')
#plt.plot(anglebins_vec, dcabins_vec, '.')
#plt.show()
plt.close()
#guess = [.1,.1,.1,.1,.1,.1]
#[a,b,c,d,e,f], params_cov = optimize.curve_fit(rb_model_func,
# va4interp, rb4interp, guess)
# #anglebins_vec, dcabins_vec, guess)
#n = Doppler(swath_files[0])
#van = np.mean(dop2correct['sensor_view'], axis=0)
#plt.plot(van, rb_model_func(van,a,b,c,d,e,f), 'r--')
#plt.plot(anglebins_vec, dcabins_vec, '.')
#plt.show()
#ww = 1./std_rb4interp
#ww[np.isinf(ww)] = 0
#rbinterp = UnivariateSpline(
# va4interp,
# rb4interp,
# w = ww,
# k = 5
# )
#van = dop2correct['sensor_view']
#y = rbinterp(van.flatten())
#rbfull = y.reshape(van.shape)
#plt.plot(np.mean(van, axis=0), np.mean(rbfull, axis=0), 'r--')
#plt.plot(anglebins_vec, dcabins_vec, '.')
#plt.show()
band_name = 'fdg_corrected'
fdg = dop2correct.anomaly() - rbfull
#plt.imshow(fdg, vmin=-60, vmax=60)
#plt.colorbar()
#plt.show()
dop2correct.add_band(array=fdg,
parameters={
'wkv':'surface_backwards_doppler_frequency_shift_of_radar_wave_due_to_surface_velocity',
'name': band_name
}
)
current = -(np.pi*(fdg - dop2correct['fww']) / 112 /
np.sin(dop2correct['incidence_angle']*np.pi/180))
dop2correct.add_band(array=current,
parameters={'name': 'current', 'units': 'm/s', 'minmax': '-2 2'}
)
land = np.array([])
# add land data for accuracy calculation
if land.shape==(0,):
land = dop2correct['valid_land_doppler'][dop2correct['valid_land_doppler'].any(axis=1)]
land_fdg = fdg[dop2correct['valid_land_doppler'].any(axis=1)]
else:
landn = dop2correct['valid_land_doppler'][dop2correct['valid_land_doppler'].any(axis=1)]
land_fdgn = fdg[dop2correct['valid_land_doppler'].any(axis=1)]
if not landn.shape==land.shape:
if landn.shape[1] > land.shape[1]:
land = np.resize(land, (land.shape[0], landn.shape[1]))
land_fdg = np.resize(land_fdg, (land_fdg.shape[0],
land_fdgn.shape[1]))
if landn.shape[1] < land.shape[1]:
landn = np.resize(landn, (landn.shape[0], land.shape[1]))
land_fdgn = np.resize(land_fdgn, (land_fdgn.shape[0],
land.shape[1]))
land = np.concatenate((land, landn))
land_fdg = np.concatenate((land_fdg, land_fdgn))
module = 'sar_doppler'
DS = Dataset.objects.get(dataseturi__uri__contains=dop2correct.fileName)
#fn = '/mnt/10.11.12.232/sat_downloads_asar/level-0/2010-01/gsar_rvl/' \
# + dop2correct.fileName.split('/')[-2]+'.gsar'
mp = media_path(module, nansat_filename( DS.dataseturi_set.get(
uri__endswith='gsar').uri))
ppath = product_path(module, nansat_filename( DS.dataseturi_set.get(
uri__endswith='gsar').uri))
# See managers.py -- this must be generalized!
pngfilename = '%s_subswath_%d.png'%(band_name, swath)
ncfilename = '%s_subswath_%d.nc'%(band_name, swath)
# Export to new netcdf with fdg as the only band
expFile = os.path.join(ppath, ncfilename)
print 'Exporting file: %s\n\n' %expFile
dop2correct.export(expFile, bands=[dop2correct._get_band_number(band_name)])
ncuri = os.path.join('file://localhost', expFile)
new_uri, created = DatasetURI.objects.get_or_create(uri=ncuri,
dataset=DS)
# Reproject to leaflet projection
xlon, xlat = dop2correct.get_corners()
dom = Domain(NSR(3857),
'-lle %f %f %f %f -tr 1000 1000' % (
xlon.min(), xlat.min(), xlon.max(), xlat.max()))
dop2correct.reproject(dom, eResampleAlg=1, tps=True)
# Update figure
dop2correct.write_figure(os.path.join(mp, pngfilename),
clim = [-60,60],
bands=band_name,
mask_array=dop2correct['swathmask'],
mask_lut={0:[128,128,128]}, transparency=[128,128,128])
print("--- %s seconds ---" % (time.time() - start_time))
land_fdg[land==0] = np.nan
print('Standard deviation over land: %.2f' %np.nanstd(land_fdg))