def get_cloud_short_variability(cloud_index, valid_measurements_mask):
"""
Wrapper returning cloud variability index computed between cloud index maps at step t and step t+1
:param cloud_index: array (slots, latitudes, longitudes) with cloud index (cli or unbiased difference)
:param valid_measurements_mask: boolean array (slots, latitudes, longitudes) indicating which measurements are valid
:return: array (slots, latitudes, longitudes) with cloud variability index between consecutive cloud index maps
"""
return compute_variability(cloud_index=cloud_index,
mask=valid_measurements_mask,
step=1)
def suspect_snow_classified_pixels(snow, ndsi, mask_input):
'''
the "suspect snow" test flagged as cloud the pixels which were supposed to be snowy
- because they pass the adequate test - but whose ndsi is strangely high
:param snow:
:param ndsi:
:param mask_input:
:return: boolean matrix (time, latitude, longitude) which is True if the pixel is snowy
'''
return snow & ((get_bright_positive_variability_5d(ndsi, mask_input, typical_time_step(), 1) > 0.2) |
(get_bright_negative_variability_5d(ndsi, mask_input, typical_time_step(), 1) > 0.2) |
(compute_variability(ndsi, mask=mask_input, abs_value=True) > 0.05))
def get_bright_positive_variability_5d(index, definition_mask, satellite_step,
slot_step):
'''
To recognize some icy clouds (peaks of ndsi)
NB: we loose information about the first slot (resp the last slot) if night is 1 slot longer during 1 of the 5 days
:param index:
:param definition_mask:
:param satellite_step: the satellite characteristic time step between two slots (10 minutes for Himawari 8)
:param slot_step: the chosen sampling of slots. if slot_step = n, the sampled slots are s[0], s[n], s[2*n]...
:return:
'''
nb_slots_per_day = get_nb_slots_per_day(satellite_step, slot_step)
nb_days = np.shape(index)[0] / nb_slots_per_day
to_return = np.full_like(index, -10)
if nb_days >= 2:
var_ndsi_1d_past = compute_variability(cloud_index=index,
mask=definition_mask,
step=nb_slots_per_day,
negative_variation_only=False,
abs_value=False)
var_ndsi_1d_future = compute_variability(cloud_index=index,
mask=definition_mask,
step=-nb_slots_per_day,
negative_variation_only=False,
abs_value=False)
if nb_days == 2:
to_return[:nb_slots_per_day] = var_ndsi_1d_future[:
nb_slots_per_day]
to_return[nb_slots_per_day:] = var_ndsi_1d_past[nb_slots_per_day:]
else: # nb_days >=3
var_ndsi_2d_past = compute_variability(
cloud_index=index,
mask=definition_mask,
step=nb_slots_per_day * 2,
negative_variation_only=False,
abs_value=False)
var_ndsi_2d_future = compute_variability(
cloud_index=index,
mask=definition_mask,
step=-2 * nb_slots_per_day,
negative_variation_only=False,
abs_value=False)
# first day
to_return[:nb_slots_per_day] = np.minimum(
var_ndsi_1d_future[:nb_slots_per_day],
var_ndsi_2d_future[:nb_slots_per_day])
# last day
to_return[-nb_slots_per_day:] = np.minimum(
var_ndsi_1d_past[-nb_slots_per_day:],
var_ndsi_2d_past[-nb_slots_per_day:])
if nb_days == 3:
# second day
to_return[nb_slots_per_day:2 * nb_slots_per_day] = np.minimum(
var_ndsi_1d_past[nb_slots_per_day:2 * nb_slots_per_day],
var_ndsi_1d_future[nb_slots_per_day:2 * nb_slots_per_day])
else: # nb_days >= 4
# the day previous the last one
to_return[-2 * nb_slots_per_day:-nb_slots_per_day] = np.minimum(
np.minimum(
var_ndsi_1d_past[-2 *
nb_slots_per_day:-nb_slots_per_day],
var_ndsi_2d_past[-2 *
nb_slots_per_day:-nb_slots_per_day]),
var_ndsi_1d_future[-2 *
nb_slots_per_day:-nb_slots_per_day])
# second day
to_return[nb_slots_per_day:2 * nb_slots_per_day] = np.minimum(
np.minimum(
var_ndsi_1d_future[nb_slots_per_day:2 *
nb_slots_per_day],
var_ndsi_2d_future[nb_slots_per_day:2 *
nb_slots_per_day]),
var_ndsi_1d_past[nb_slots_per_day:2 * nb_slots_per_day])
if nb_days >= 5:
to_return[2 * nb_slots_per_day:-2 *
nb_slots_per_day] = np.minimum(
np.minimum(
var_ndsi_1d_past[2 *
nb_slots_per_day:-2 *
nb_slots_per_day],
var_ndsi_2d_past[2 *
nb_slots_per_day:-2 *
nb_slots_per_day]),
np.minimum(
var_ndsi_1d_future[2 *
nb_slots_per_day:-2 *
nb_slots_per_day],
var_ndsi_2d_future[2 *
nb_slots_per_day:-2 *
nb_slots_per_day]))
to_return[to_return < 0] = 0
return to_return
def get_cloud_index_total_variability_7d(cloud_index, definition_mask,
pre_cloud_mask, satellite_step,
slot_step):
'''
:param cloud_index: array (slots, latitudes, longitudes) with cloud index (cli or unbiased difference)
:param definition_mask: mask points where cloud index is not defined
:param pre_cloud_mask: mask points where we don't want to compute 5 days variability (eg: obvious water clouds)
:param satellite_step: the satellite characteristic time step between two slots (10 minutes for Himawari 8)
:param slot_step: the chosen sampling of slots. if slot_step = n, the sampled slots are s[0], s[n], s[2*n]...
:return:
'''
if pre_cloud_mask is not None:
mask = definition_mask | pre_cloud_mask
else:
mask = definition_mask
nb_slots_per_day = get_nb_slots_per_day(satellite_step, slot_step)
nb_days = np.shape(cloud_index)[0] / nb_slots_per_day
to_return = np.full_like(cloud_index, -10)
if nb_days >= 2:
var_cli_1d_past = compute_variability(cloud_index=cloud_index,
mask=mask,
abs_value=True,
step=nb_slots_per_day)
var_cli_1d_future = compute_variability(cloud_index=cloud_index,
mask=mask,
abs_value=True,
step=-nb_slots_per_day)
if nb_days == 2:
to_return[:nb_slots_per_day] = var_cli_1d_future[:nb_slots_per_day]
to_return[nb_slots_per_day:] = var_cli_1d_past[nb_slots_per_day:]
else: # nb_days >=3
var_cli_2d_past = compute_variability(cloud_index=cloud_index,
mask=mask,
abs_value=True,
step=nb_slots_per_day * 2)
var_cli_2d_future = compute_variability(cloud_index=cloud_index,
mask=mask,
abs_value=True,
step=-2 * nb_slots_per_day)
# first day
to_return[:nb_slots_per_day] = np.minimum(
var_cli_1d_future[:nb_slots_per_day],
var_cli_2d_future[:nb_slots_per_day])
# last day
to_return[-nb_slots_per_day:] = np.minimum(
var_cli_1d_past[-nb_slots_per_day:],
var_cli_2d_past[-nb_slots_per_day:])
if nb_days == 3:
# second day
to_return[nb_slots_per_day:2 * nb_slots_per_day] = np.minimum(
var_cli_1d_past[nb_slots_per_day:2 * nb_slots_per_day],
var_cli_1d_future[nb_slots_per_day:2 * nb_slots_per_day])
else: # nb_days >= 4
# the day previous the last one
to_return[-2 * nb_slots_per_day:-nb_slots_per_day] = np.minimum(
np.minimum(
var_cli_1d_past[-2 *
nb_slots_per_day:-nb_slots_per_day],
var_cli_2d_past[-2 *
nb_slots_per_day:-nb_slots_per_day]),
var_cli_1d_future[-2 * nb_slots_per_day:-nb_slots_per_day])
# second day
to_return[nb_slots_per_day:2 * nb_slots_per_day] = np.minimum(
np.minimum(
var_cli_1d_future[nb_slots_per_day:2 *
nb_slots_per_day],
var_cli_2d_future[nb_slots_per_day:2 *
nb_slots_per_day]),
var_cli_1d_past[nb_slots_per_day:2 * nb_slots_per_day])
if nb_days == 5:
to_return[2 * nb_slots_per_day:-2 *
nb_slots_per_day] = np.minimum(
np.minimum(
var_cli_1d_past[2 * nb_slots_per_day:-2 *
nb_slots_per_day],
var_cli_2d_past[2 * nb_slots_per_day:-2 *
nb_slots_per_day]),
np.minimum(
var_cli_1d_future[2 *
nb_slots_per_day:-2 *
nb_slots_per_day],
var_cli_2d_future[2 *
nb_slots_per_day:-2 *
nb_slots_per_day]))
else: # nb_days >= 6
var_cli_3d_past = compute_variability(
cloud_index=cloud_index,
mask=mask,
abs_value=True,
step=nb_slots_per_day * 3)
var_cli_3d_future = compute_variability(
cloud_index=cloud_index,
mask=mask,
abs_value=True,
step=-nb_slots_per_day * 3)
# two days previous the last one
to_return[-3 * nb_slots_per_day:-2 *
nb_slots_per_day] = np.minimum(
np.minimum(
np.minimum(
var_cli_1d_past[-3 *
nb_slots_per_day:-2 *
nb_slots_per_day],
var_cli_2d_past[-3 *
nb_slots_per_day:-2 *
nb_slots_per_day]),
var_cli_3d_past[-3 *
nb_slots_per_day:-2 *
nb_slots_per_day]),
np.minimum(
var_cli_1d_future[-3 *
nb_slots_per_day:-2 *
nb_slots_per_day],
var_cli_2d_future[-3 *
nb_slots_per_day:-2 *
nb_slots_per_day]))
# third day
to_return[
2 * nb_slots_per_day:3 *
nb_slots_per_day] = np.minimum(
np.minimum(
np.minimum(
var_cli_1d_future[2 * nb_slots_per_day:3 *
nb_slots_per_day],
var_cli_2d_future[2 * nb_slots_per_day:3 *
nb_slots_per_day]),
var_cli_3d_future[2 * nb_slots_per_day:3 *
nb_slots_per_day]),
np.minimum(
var_cli_1d_past[2 * nb_slots_per_day:3 *
nb_slots_per_day],
var_cli_2d_past[2 * nb_slots_per_day:3 *
nb_slots_per_day]))
if nb_days >= 7:
to_return[
3 * nb_slots_per_day:-3 *
nb_slots_per_day] = np.minimum(
np.minimum(
np.minimum(
var_cli_1d_past[3 *
nb_slots_per_day:-3 *
nb_slots_per_day],
var_cli_2d_past[3 *
nb_slots_per_day:-3 *
nb_slots_per_day]),
var_cli_3d_past[3 * nb_slots_per_day:-3 *
nb_slots_per_day]),
np.minimum(
np.minimum(
var_cli_1d_future[3 *
nb_slots_per_day:-3 *
nb_slots_per_day],
var_cli_2d_future[3 *
nb_slots_per_day:-3 *
nb_slots_per_day]),
var_cli_3d_future[3 * nb_slots_per_day:-3 *
nb_slots_per_day]))
return to_return
def get_cloud_index_positive_variability_7d(cloud_index,
definition_mask,
satellite_step,
pre_cloud_mask=None,
slot_step=1,
only_past=False):
'''
This function is supposed to help finding small clouds with low positive clouds (in partiular NOT icy clouds)
:param only_past: MODE of the function. If True, compare one day with the previous days only.
If False, compare one day both with the previous and the following days.
:param cloud_index: array (slots, latitudes, longitudes) with cloud index (cli or unbiased difference)
:param definition_mask: mask points where cloud index is not defined
:param pre_cloud_mask: mask points where we don't want to compute 5 days variability (eg: obvious water clouds)
:param satellite_step: the satellite characteristic time step between two slots (10 minutes for Himawari 8)
:param slot_step: the chosen sampling of slots. if slot_step = n, the sampled slots are s[0], s[n], s[2*n]...
:return:
'''
if pre_cloud_mask is not None:
mask = definition_mask | pre_cloud_mask
else:
mask = definition_mask
slots_per_day = get_nb_slots_per_day(satellite_step, slot_step)
nb_days = np.shape(cloud_index)[0] / slots_per_day
to_return = np.zeros_like(cloud_index)
if only_past:
if nb_days >= 2:
var_cli_1d_past = compute_variability(cloud_index=cloud_index,
mask=mask,
step=slots_per_day)
to_return[slots_per_day:] = var_cli_1d_past[slots_per_day:]
if nb_days >= 3:
var_cli_2d_past = compute_variability(cloud_index=cloud_index,
mask=mask,
step=slots_per_day * 2)
# second day
to_return[slots_per_day:2 * slots_per_day] = np.maximum(
var_cli_1d_past[slots_per_day:2 * slots_per_day],
var_cli_2d_past[-slots_per_day:2 * slots_per_day])
if nb_days == 3:
# second day
to_return[slots_per_day:2 * slots_per_day] = np.maximum(
var_cli_1d_past[slots_per_day:2 * slots_per_day],
var_cli_1d_future[slots_per_day:2 * slots_per_day])
else: # nb_days >= 4
# the day previous the last one
to_return[-2 * slots_per_day:-slots_per_day] = np.maximum(
np.maximum(
var_cli_1d_past[-2 * slots_per_day:-slots_per_day],
var_cli_2d_past[-2 *
slots_per_day:-slots_per_day]),
var_cli_1d_future[-2 * slots_per_day:-slots_per_day])
# second day
to_return[slots_per_day:2 * slots_per_day] = np.maximum(
np.maximum(
var_cli_1d_future[slots_per_day:2 * slots_per_day],
var_cli_2d_future[slots_per_day:2 *
slots_per_day]),
var_cli_1d_past[slots_per_day:2 * slots_per_day])
if nb_days >= 5:
to_return[2 * slots_per_day:-2 *
slots_per_day] = np.maximum(
np.maximum(
var_cli_1d_past[2 *
slots_per_day:-2 *
slots_per_day],
var_cli_2d_past[2 *
slots_per_day:-2 *
slots_per_day]),
np.maximum(
var_cli_1d_future[2 *
slots_per_day:-2 *
slots_per_day],
var_cli_2d_future[2 *
slots_per_day:-2 *
slots_per_day]))
else: # nb_days >= 6
var_cli_3d_past = compute_variability(
cloud_index=cloud_index,
mask=mask,
step=slots_per_day * 3)
var_cli_3d_future = compute_variability(
cloud_index=cloud_index,
mask=mask,
step=-slots_per_day * 3)
# two days previous the last one
to_return[
-3 * slots_per_day:-2 *
slots_per_day] = np.maximum(
np.maximum(
np.maximum(
var_cli_1d_past[-3 * slots_per_day:-2 *
slots_per_day],
var_cli_2d_past[-3 * slots_per_day:-2 *
slots_per_day]),
var_cli_3d_past[-3 * slots_per_day:-2 *
slots_per_day]),
np.maximum(
var_cli_1d_future[-3 * slots_per_day:-2 *
slots_per_day],
var_cli_2d_future[-3 * slots_per_day:-2 *
slots_per_day]))
# third day
to_return[
2 * slots_per_day:3 * slots_per_day] = np.maximum(
np.maximum(
np.maximum(
var_cli_1d_future[2 * slots_per_day:3 *
slots_per_day],
var_cli_2d_future[2 * slots_per_day:3 *
slots_per_day]),
var_cli_3d_future[2 * slots_per_day:3 *
slots_per_day]),
np.maximum(
var_cli_1d_past[2 * slots_per_day:3 *
slots_per_day],
var_cli_2d_past[2 * slots_per_day:3 *
slots_per_day]))
if nb_days >= 7:
to_return[
3 * slots_per_day:-3 *
slots_per_day] = np.maximum(
np.maximum(
np.maximum(
var_cli_1d_past[3 *
slots_per_day:-3 *
slots_per_day],
var_cli_2d_past[3 *
slots_per_day:-3 *
slots_per_day]),
var_cli_3d_past[3 * slots_per_day:-3 *
slots_per_day]),
np.maximum(
np.maximum(
var_cli_1d_future[
3 * slots_per_day:-3 *
slots_per_day],
var_cli_2d_future[
3 * slots_per_day:-3 *
slots_per_day]),
var_cli_3d_future[3 *
slots_per_day:-3 *
slots_per_day]))
else:
if nb_days >= 2:
var_cli_1d_past = compute_variability(cloud_index=cloud_index,
mask=mask,
step=slots_per_day)
var_cli_1d_future = compute_variability(cloud_index=cloud_index,
mask=mask,
step=-slots_per_day)
if nb_days == 2:
to_return[:slots_per_day] = var_cli_1d_future[:slots_per_day]
to_return[slots_per_day:] = var_cli_1d_past[slots_per_day:]
else: # nb_days >=3
var_cli_2d_past = compute_variability(cloud_index=cloud_index,
mask=mask,
step=slots_per_day * 2)
var_cli_2d_future = compute_variability(
cloud_index=cloud_index,
mask=mask,
step=-2 * slots_per_day)
# first day
to_return[:slots_per_day] = np.maximum(
var_cli_1d_future[:slots_per_day],
var_cli_2d_future[:slots_per_day])
# last day
to_return[-slots_per_day:] = np.maximum(
var_cli_1d_past[-slots_per_day:],
var_cli_2d_past[-slots_per_day:])
if nb_days == 3:
# second day
to_return[slots_per_day:2 * slots_per_day] = np.maximum(
var_cli_1d_past[slots_per_day:2 * slots_per_day],
var_cli_1d_future[slots_per_day:2 * slots_per_day])
else: # nb_days >= 4
# the day previous the last one
to_return[-2 * slots_per_day:-slots_per_day] = np.maximum(
np.maximum(
var_cli_1d_past[-2 * slots_per_day:-slots_per_day],
var_cli_2d_past[-2 *
slots_per_day:-slots_per_day]),
var_cli_1d_future[-2 * slots_per_day:-slots_per_day])
# second day
to_return[slots_per_day:2 * slots_per_day] = np.maximum(
np.maximum(
var_cli_1d_future[slots_per_day:2 * slots_per_day],
var_cli_2d_future[slots_per_day:2 *
slots_per_day]),
var_cli_1d_past[slots_per_day:2 * slots_per_day])
if nb_days >= 5:
to_return[2 * slots_per_day:-2 *
slots_per_day] = np.maximum(
np.maximum(
var_cli_1d_past[2 *
slots_per_day:-2 *
slots_per_day],
var_cli_2d_past[2 *
slots_per_day:-2 *
slots_per_day]),
np.maximum(
var_cli_1d_future[2 *
slots_per_day:-2 *
slots_per_day],
var_cli_2d_future[2 *
slots_per_day:-2 *
slots_per_day]))
else: # nb_days >= 6
var_cli_3d_past = compute_variability(
cloud_index=cloud_index,
mask=mask,
step=slots_per_day * 3)
var_cli_3d_future = compute_variability(
cloud_index=cloud_index,
mask=mask,
step=-slots_per_day * 3)
# two days previous the last one
to_return[
-3 * slots_per_day:-2 *
slots_per_day] = np.maximum(
np.maximum(
np.maximum(
var_cli_1d_past[-3 * slots_per_day:-2 *
slots_per_day],
var_cli_2d_past[-3 * slots_per_day:-2 *
slots_per_day]),
var_cli_3d_past[-3 * slots_per_day:-2 *
slots_per_day]),
np.maximum(
var_cli_1d_future[-3 * slots_per_day:-2 *
slots_per_day],
var_cli_2d_future[-3 * slots_per_day:-2 *
slots_per_day]))
# third day
to_return[
2 * slots_per_day:3 * slots_per_day] = np.maximum(
np.maximum(
np.maximum(
var_cli_1d_future[2 * slots_per_day:3 *
slots_per_day],
var_cli_2d_future[2 * slots_per_day:3 *
slots_per_day]),
var_cli_3d_future[2 * slots_per_day:3 *
slots_per_day]),
np.maximum(
var_cli_1d_past[2 * slots_per_day:3 *
slots_per_day],
var_cli_2d_past[2 * slots_per_day:3 *
slots_per_day]))
if nb_days >= 7:
to_return[
3 * slots_per_day:-3 *
slots_per_day] = np.maximum(
np.maximum(
np.maximum(
var_cli_1d_past[3 *
slots_per_day:-3 *
slots_per_day],
var_cli_2d_past[3 *
slots_per_day:-3 *
slots_per_day]),
var_cli_3d_past[3 * slots_per_day:-3 *
slots_per_day]),
np.maximum(
np.maximum(
var_cli_1d_future[
3 * slots_per_day:-3 *
slots_per_day],
var_cli_2d_future[
3 * slots_per_day:-3 *
slots_per_day]),
var_cli_3d_future[3 *
slots_per_day:-3 *
slots_per_day]))
# we are interested only in positive cli variations
to_return[to_return < 0] = 0
return to_return