def command(paths):
""" Show a lowest elevation image. """
# order
index = dict(
NL60=0,
NL61=1,
nhb=2,
ess=3,
emd=4,
JAB=5,
)
# load
d_rang, d_elev, d_anth = {}, {}, {}
for p in paths:
with h5py.File(p, 'r') as h5:
for r in h5:
if r in d_rang or r == 'ase':
continue
d_rang[r] = h5[r]['range'][:]
d_elev[r] = h5[r]['elevation'][:]
d_anth[r] = h5[r].attrs['antenna_height']
radars = d_anth.keys()
elev = np.ma.empty((len(radars), ) + d_rang[radars[0]].shape)
rang = np.ma.empty((len(radars), ) + d_rang[radars[0]].shape)
anth = np.empty((len(radars), 1, 1))
for r in radars:
elev[index[r]] = np.ma.masked_equal(d_elev[r], config.NODATAVALUE)
rang[index[r]] = np.ma.masked_equal(d_rang[r], config.NODATAVALUE)
anth[index[r]] = float(d_anth[r]) / 1000
# calculate
theta = calc.calculate_theta(
rang=rang,
elev=np.radians(elev),
anth=anth,
)
alt = calc.calculate_height(
theta=theta,
elev=np.radians(elev),
anth=anth,
)
which = np.ma.where(
alt == alt.min(0),
np.indices(alt.shape)[0],
-1,
).max(0)
what = alt.min(0)
# colors
hue = cm.hsv(colors.Normalize(vmax=len(radars))(which), bytes=True)
sat = 1 - colors.Normalize(vmax=5, clip=True)(what)[..., np.newaxis]
hue[..., :3] *= sat
hue[sat.mask[..., 0]] = 255
Image.fromarray(hue).save('elevation_image.png')
return 0
def command(paths):
""" Show a lowest elevation image. """
# order
index = dict(
NL60=0,
NL61=1,
nhb=2,
ess=3,
emd=4,
JAB=5,
)
# load
d_rang, d_elev, d_anth = {}, {}, {}
for p in paths:
with h5py.File(p, 'r') as h5:
for r in h5:
if r in d_rang or r == 'ase':
continue
d_rang[r] = h5[r]['range'][:]
d_elev[r] = h5[r]['elevation'][:]
d_anth[r] = h5[r].attrs['antenna_height']
radars = d_anth.keys()
elev = np.ma.empty((len(radars),) + d_rang[radars[0]].shape)
rang = np.ma.empty((len(radars),) + d_rang[radars[0]].shape)
anth = np.empty((len(radars), 1, 1))
for r in radars:
elev[index[r]] = np.ma.masked_equal(d_elev[r], config.NODATAVALUE)
rang[index[r]] = np.ma.masked_equal(d_rang[r], config.NODATAVALUE)
anth[index[r]] = float(d_anth[r]) / 1000
# calculate
theta = calc.calculate_theta(
rang=rang, elev=np.radians(elev), anth=anth,
)
alt = calc.calculate_height(
theta=theta, elev=np.radians(elev), anth=anth,
)
which = np.ma.where(
alt == alt.min(0),
np.indices(alt.shape)[0],
-1,
).max(0)
what = alt.min(0)
# colors
hue = cm.hsv(colors.Normalize(vmax=len(radars))(which), bytes=True)
sat = 1 - colors.Normalize(vmax=5, clip=True)(what)[..., np.newaxis]
hue[..., :3] *= sat
hue[sat.mask[..., 0]] = 255
Image.fromarray(hue).save('elevation_image.png')
return 0
def _calculate(self, datasets):
"""
Return a composite dataset based on the
weighted lowest altitudes method.
"""
if not datasets:
return np.ma.array(
np.zeros(self.grid.get_shape()),
mask=True,
fill_value=config.NODATAVALUE,
)
# Read datasets
metadata = [dict(ds.attrs) for ds in datasets]
stations = [m['station'] for m in metadata]
anth = np.array(
[json.loads(m['antenna_height']) for m in metadata],
).reshape((-1, 1, 1)) / 1000
rain = np.ma.array(
[gridtools.h5ds2ma(ds['rain']) for ds in datasets],
fill_value=config.NODATAVALUE,
)
rang = np.ma.array(
[gridtools.h5ds2ma(ds['range']) for ds in datasets],
fill_value=config.NODATAVALUE,
)
elev = np.ma.array(
[gridtools.h5ds2ma(ds['elevation']) for ds in datasets],
fill_value=config.NODATAVALUE,
)
# Extend mask around NL stations
if 'NL61' in stations and 'NL62' in stations:
for i in map(stations.index, ['NL61', 'NL62']):
rain.mask[i][np.less(rang[i], 15)] = True
# Extend mask around JAB
if 'JAB' in stations and 'NL62' in stations:
i = stations.index('JAB')
rain.mask[i][np.less(rang[i], 15)] = True
# Do history declutter
if self.declutter['history']: # None or 0 disables history declutter
logging.debug('Starting history declutter, threshold {}'.format(
self.declutter['history']
))
# Initialize clutter array of same dimensions as rain array
clutter = np.zeros(rain.shape, rain.dtype)
declutter_mask = h5py.File(
os.path.join(config.MISC_DIR, config.DECLUTTER_FILEPATH), 'r',
)
left, right, upper, lower = self._get_window(self.grid)
for i, radar in enumerate(stations):
if radar in declutter_mask:
clutter[i, upper:lower, left:right] = declutter_mask[radar]
declutter_mask.close()
while True:
clutter[rain.mask] = 0
extra = reduce(np.logical_and, [
# clutter above threshold for that pixel
np.greater(clutter, self.declutter['history']),
# at least two unmasked pixels left
np.greater((~rain.mask).sum(0), 1),
# the maximum clutter must be masked
np.equal(clutter, clutter.max(0)),
])
# Extend rain mask with cluttermask.
count = extra.sum()
if not count:
break
logging.debug(
'Masking {} historically suspicious pixels'.format(count),
)
rain.mask[extra] = True
rang.mask = rain.mask
elev.mask = rain.mask
# Calculate heights
theta = calc.calculate_theta(
rang=rang, elev=np.radians(elev), anth=anth,
)
alt_min = calc.calculate_height(
theta=theta, elev=np.radians(elev - 0.5), anth=anth,
)
alt_max = calc.calculate_height(
theta=theta, elev=np.radians(elev + 0.5), anth=anth,
)
lowest_alt_max = np.ma.where(
np.equal(alt_min, alt_min.min(0)),
alt_max,
np.ma.masked,
).min(0)
bandwidth = alt_max - alt_min
vertdist1 = (lowest_alt_max - alt_min)
vertdist1[np.ma.less(vertdist1, 0)] = 0
vertdist2 = (lowest_alt_max - alt_max)
vertdist2[np.ma.less(vertdist2, 0)] = 0
overlap = vertdist1 - vertdist2
weight = overlap / bandwidth
composite = (rain * weight).sum(0) / weight.sum(0)
composite.fill_value = config.NODATAVALUE
return composite
def get(self):
""" Return gdal dataset in rd projection. """
data = self.data()
rang, azim, elev = data['polar']
rain = data['rain']
latlon = data['latlon']
anth = data['ant_alt']
theta = calc.calculate_theta(
rang=rang,
elev=np.radians(elev),
anth=anth / 1000,
)
points_aeqd = calc.calculate_cartesian(
theta=theta,
azim=np.radians(azim),
)
projections = (
utils.projection_aeqd(*latlon),
utils.projection(utils.RD),
)
points_rd = utils.coordinate_transformer.transform(
points=points_aeqd,
projections=projections,
)
# Interpolate
grid_rain, grid_rang, grid_elev = self._interpolate(
points=points_rd,
values=(
rain,
rang * np.ones(rain.shape),
elev * np.ones(rain.shape),
),
grid=self.grid.get_grid(),
)
location = utils.transform((0, 0), projections)
ds_rd = self.grid.create_dataset(bands=3)
ds_rd.SetMetadata(dict(
source=self.signature.get_scanname(),
timestamp=self.signature.get_timestamp(),
station=self.signature.get_code(),
location=json.dumps(location),
antenna_height=json.dumps(anth),
max_elevation=json.dumps(elev.max()),
min_elevation=json.dumps(elev.min()),
max_range=json.dumps(rang.max()),
))
banddata = {
BAND_RAIN: grid_rain,
BAND_RANG: grid_rang,
BAND_ELEV: grid_elev,
}
for i in range(1, ds_rd.RasterCount + 1):
band = ds_rd.GetRasterBand(i)
band.SetNoDataValue(banddata[i].fill_value)
band.WriteArray(banddata[i].filled())
band.SetMetadata(BAND_META[i])
return ds_rd
def _calculate(self, datasets):
"""
Return a composite dataset based on the
weighted lowest altitudes method.
"""
if not datasets:
return np.ma.array(
np.zeros(self.grid.get_shape()),
mask=True,
fill_value=config.NODATAVALUE,
)
# Read datasets
metadata = [dict(ds.attrs) for ds in datasets]
stations = [m['station'] for m in metadata]
anth = np.array([json.loads(m['antenna_height'])
for m in metadata], ).reshape((-1, 1, 1)) / 1000
rain = np.ma.array(
[gridtools.h5ds2ma(ds['rain']) for ds in datasets],
fill_value=config.NODATAVALUE,
)
rang = np.ma.array(
[gridtools.h5ds2ma(ds['range']) for ds in datasets],
fill_value=config.NODATAVALUE,
)
elev = np.ma.array(
[gridtools.h5ds2ma(ds['elevation']) for ds in datasets],
fill_value=config.NODATAVALUE,
)
# Extend mask around NL stations
if 'NL60' in stations and 'NL61' in stations:
for i in map(stations.index, ['NL60', 'NL61']):
rain.mask[i][np.less(rang[i], 15)] = True
# Extend mask around JAB
if 'JAB' in stations and 'NL60' in stations:
i = stations.index('JAB')
rain.mask[i][np.less(rang[i], 15)] = True
# Do history declutter
if self.declutter['history']: # None or 0 disables history declutter
logging.debug('Starting history declutter, threshold {}'.format(
self.declutter['history']))
# Initialize clutter array of same dimensions as rain array
clutter = np.zeros(rain.shape, rain.dtype)
declutter_mask = h5py.File(
os.path.join(config.MISC_DIR, config.DECLUTTER_FILEPATH),
'r',
)
left, right, upper, lower = self._get_window(self.grid)
for i, radar in enumerate(stations):
if radar in declutter_mask:
clutter[i, upper:lower, left:right] = declutter_mask[radar]
declutter_mask.close()
while True:
clutter[rain.mask] = 0
extra = reduce(
np.logical_and,
[
# clutter above threshold for that pixel
np.greater(clutter, self.declutter['history']),
# at least two unmasked pixels left
np.greater((~rain.mask).sum(0), 1),
# the maximum clutter must be masked
np.equal(clutter, clutter.max(0)),
])
# Extend rain mask with cluttermask.
count = extra.sum()
if not count:
break
logging.debug(
'Masking {} historically suspicious pixels'.format(count),
)
rain.mask[extra] = True
rang.mask = rain.mask
elev.mask = rain.mask
# Calculate heights
theta = calc.calculate_theta(
rang=rang,
elev=np.radians(elev),
anth=anth,
)
alt_min = calc.calculate_height(
theta=theta,
elev=np.radians(elev - 0.5),
anth=anth,
)
alt_max = calc.calculate_height(
theta=theta,
elev=np.radians(elev + 0.5),
anth=anth,
)
lowest_alt_max = np.ma.where(
np.equal(alt_min, alt_min.min(0)),
alt_max,
np.ma.masked,
).min(0)
bandwidth = alt_max - alt_min
vertdist1 = (lowest_alt_max - alt_min)
vertdist1[np.ma.less(vertdist1, 0)] = 0
vertdist2 = (lowest_alt_max - alt_max)
vertdist2[np.ma.less(vertdist2, 0)] = 0
overlap = vertdist1 - vertdist2
weight = overlap / bandwidth
composite = (rain * weight).sum(0) / weight.sum(0)
composite.fill_value = config.NODATAVALUE
return composite
def get(self):
""" Return gdal dataset in rd projection. """
data = self.data()
rang, azim, elev = data['polar']
rain = data['rain']
latlon = data['latlon']
anth = data['ant_alt']
theta = calc.calculate_theta(
rang=rang,
elev=np.radians(elev),
anth=anth / 1000,
)
points_aeqd = calc.calculate_cartesian(
theta=theta,
azim=np.radians(azim),
)
projections = (
utils.projection_aeqd(*latlon),
utils.projection(utils.RD),
)
points_rd = utils.coordinate_transformer.transform(
points=points_aeqd,
projections=projections,
)
# Interpolate
grid_rain, grid_rang, grid_elev = self._interpolate(
points=points_rd,
values=(
rain,
rang * np.ones(rain.shape),
elev * np.ones(rain.shape),
),
grid=self.grid.get_grid(),
)
location = utils.transform((0, 0), projections)
ds_rd = self.grid.create_dataset(bands=3)
ds_rd.SetMetadata(
dict(
source=self.signature.get_scanname(),
timestamp=self.signature.get_timestamp(),
station=self.signature.get_code(),
location=json.dumps(location),
antenna_height=json.dumps(anth),
max_elevation=json.dumps(elev.max()),
min_elevation=json.dumps(elev.min()),
max_range=json.dumps(rang.max()),
))
banddata = {
BAND_RAIN: grid_rain,
BAND_RANG: grid_rang,
BAND_ELEV: grid_elev,
}
for i in range(1, ds_rd.RasterCount + 1):
band = ds_rd.GetRasterBand(i)
band.SetNoDataValue(banddata[i].fill_value)
band.WriteArray(banddata[i].filled())
band.SetMetadata(BAND_META[i])
return ds_rd