def find_model_patch_tracks(self):
"""
Identify storms in gridded model output and extract uniform sized patches around the storm centers of mass.
Returns:
"""
self.model_grid.load_data()
tracked_model_objects = []
model_objects = []
if self.model_grid.data is None:
print("No model output found")
return tracked_model_objects
min_orig = self.model_ew.min_thresh
max_orig = self.model_ew.max_thresh
data_increment_orig = self.model_ew.data_increment
self.model_ew.min_thresh = 0
self.model_ew.data_increment = 1
self.model_ew.max_thresh = 100
for h, hour in enumerate(self.hours):
# Identify storms at each time step and apply size filter
print("Finding {0} objects for run {1} Hour: {2:02d}".format(self.ensemble_member,
self.run_date.strftime("%Y%m%d%H"), hour))
if self.mask is not None:
model_data = self.model_grid.data[h] * self.mask
else:
model_data = self.model_grid.data[h]
model_data[:self.patch_radius] = 0
model_data[-self.patch_radius:] = 0
model_data[:, :self.patch_radius] = 0
model_data[:, -self.patch_radius:] = 0
scaled_data = np.array(rescale_data(model_data, min_orig, max_orig))
hour_labels = label_storm_objects(scaled_data, "ew",
self.model_ew.min_thresh, self.model_ew.max_thresh,
min_area=self.size_filter, max_area=self.model_ew.max_size,
max_range=self.model_ew.delta, increment=self.model_ew.data_increment,
gaussian_sd=self.gaussian_window)
model_objects.extend(extract_storm_patches(hour_labels, model_data, self.model_grid.x,
self.model_grid.y, [hour],
dx=self.model_grid.dx,
patch_radius=self.patch_radius))
for model_obj in model_objects[-1]:
dims = model_obj.timesteps[-1].shape
if h > 0:
model_obj.estimate_motion(hour, self.model_grid.data[h-1], dims[1], dims[0])
del scaled_data
del model_data
del hour_labels
tracked_model_objects.extend(track_storms(model_objects, self.hours,
self.object_matcher.cost_function_components,
self.object_matcher.max_values,
self.object_matcher.weights))
self.model_ew.min_thresh = min_orig
self.model_ew.max_thresh = max_orig
self.model_ew.data_increment = data_increment_orig
return tracked_model_objects
def find_model_patch_tracks(self):
"""
Identify storms in gridded model output and extract uniform sized patches around the storm centers of mass.
Returns:
"""
self.model_grid.load_data()
tracked_model_objects = []
model_objects = []
if self.model_grid.data is None:
print("No model output found")
return tracked_model_objects
min_orig = self.model_ew.min_thresh
max_orig = self.model_ew.max_thresh
data_increment_orig = self.model_ew.data_increment
self.model_ew.min_thresh = 0
self.model_ew.data_increment = 1
self.model_ew.max_thresh = 100
for h, hour in enumerate(self.hours):
# Identify storms at each time step and apply size filter
print("Finding {0} objects for run {1} Hour: {2:02d}".format(self.ensemble_member,
self.run_date.strftime("%Y%m%d%H"), hour))
if self.mask is not None:
model_data = self.model_grid.data[h] * self.mask
else:
model_data = self.model_grid.data[h]
model_data[:self.patch_radius] = 0
model_data[-self.patch_radius:] = 0
model_data[:, :self.patch_radius] = 0
model_data[:, -self.patch_radius:] = 0
scaled_data = np.array(rescale_data(model_data, min_orig, max_orig))
hour_labels = label_storm_objects(scaled_data, "ew",
self.model_ew.min_thresh, self.model_ew.max_thresh,
min_area=self.size_filter, max_area=self.model_ew.max_size,
max_range=self.model_ew.delta, increment=self.model_ew.data_increment,
gaussian_sd=self.gaussian_window)
model_objects.extend(extract_storm_patches(hour_labels, model_data, self.model_grid.x,
self.model_grid.y, [hour],
dx=self.model_grid.dx,
patch_radius=self.patch_radius))
for model_obj in model_objects[-1]:
dims = model_obj.timesteps[-1].shape
if h > 0:
model_obj.estimate_motion(hour, self.model_grid.data[h-1], dims[1], dims[0])
del scaled_data
del model_data
del hour_labels
tracked_model_objects.extend(track_storms(model_objects, self.hours,
self.object_matcher.cost_function_components,
self.object_matcher.max_values,
self.object_matcher.weights))
self.model_ew.min_thresh = min_orig
self.model_ew.max_thresh = max_orig
self.model_ew.data_increment = data_increment_orig
return tracked_model_objects
def create_patches_hourly(self, num, data, lats, lons, thetimes,
times_thisfile):
"""Function to find storm patches in WRF CONUS1 dataset. Saves output to Xarray dataset with metadata.
Args:
num (int): Number of job assignment (enumerated loop).
data (numpy array): Dbz data to use for storm patch extraction.
lats (numpy array): Latitudes of dbz data being used for storm patch extraction.
lons (numpy array): Longitudes of dbz data being used for storm patch extraction.
thetimes (numpy array): Time indices of the full time period of the climate simulations (2000-2013).
times_thisfile (DatetimeIndex): Pandas date range.
Returns:
num (int): Number of job assignment (enumerated loop).
"""
thelabels = label_storm_objects(data,
method=self.method,
min_intensity=self.min_dbz,
max_intensity=self.max_dbz,
min_area=1,
max_area=100,
max_range=1,
increment=1,
gaussian_sd=0)
print(num, "Postlabel")
storm_objs = extract_storm_patches(label_grid=thelabels,
data=data,
x_grid=lons,
y_grid=lats,
times=thetimes,
dx=1,
dt=1,
patch_radius=self.patch_radius)
print(num, f"Done {times_thisfile[num].strftime('%Y-%m-%d')}")
data_assemble = xr.Dataset(
{
'grid':
(['starttime', 'y', 'x'],
np.array([
other.timesteps[0] for obj in storm_objs
for other in obj if other.timesteps[0].shape[0] *
other.timesteps[0].shape[1] == self.total_pixels()
])),
'mask':
(['starttime', 'y', 'x'],
np.array([
other.masks[0] for obj in storm_objs
for other in obj if other.timesteps[0].shape[0] *
other.timesteps[0].shape[1] == self.total_pixels()
])),
'row_indices':
(['starttime', 'y', 'x'],
np.array([
other.i[0] for obj in storm_objs
for other in obj if other.timesteps[0].shape[0] *
other.timesteps[0].shape[1] == self.total_pixels()
])),
'col_indices':
(['starttime', 'y', 'x'],
np.array([
other.j[0] for obj in storm_objs
for other in obj if other.timesteps[0].shape[0] *
other.timesteps[0].shape[1] == self.total_pixels()
])),
'lats':
(['starttime', 'y', 'x'],
np.array([
other.y[0] for obj in storm_objs
for other in obj if other.timesteps[0].shape[0] *
other.timesteps[0].shape[1] == self.total_pixels()
])),
'lons':
(['starttime', 'y', 'x'],
np.array([
other.x[0] for obj in storm_objs
for other in obj if other.timesteps[0].shape[0] *
other.timesteps[0].shape[1] == self.total_pixels()
])),
},
coords={
'starttime':
(['starttime'],
np.array([
other.start_time for obj in storm_objs for other in obj
if other.timesteps[0].shape[0] *
other.timesteps[0].shape[1] == self.total_pixels()
])),
'endtime':
np.array([
other.end_time for obj in storm_objs for other in obj
if other.timesteps[0].shape[0] *
other.timesteps[0].shape[1] == self.total_pixels()
]),
'x_speed':
(['starttime'],
np.array([
other.u[0]
for obj in storm_objs for other in obj
if other.timesteps[0].shape[0] *
other.timesteps[0].shape[1] == self.total_pixels()
])),
'y_speed':
(['starttime'],
np.array([
other.v[0]
for obj in storm_objs for other in obj
if other.timesteps[0].shape[0] *
other.timesteps[0].shape[1] == self.total_pixels()
]))
})
data_assemble.to_netcdf(
f"/{self.destination_path}/{self.climate}_SPhourly_{times_thisfile[num].strftime('%Y%m%d')}.nc"
)
return (num)
