def _plot_saliency_for_2d3d_radar(
list_of_input_matrices, list_of_saliency_matrices,
training_option_dict, saliency_colour_map_object,
max_colour_value_by_example, output_dir_name, storm_ids=None,
storm_times_unix_sec=None):
"""Plots saliency for 2-D azimuthal-shear and 3-D reflectivity fields.
E = number of examples (storm objects)
If `storm_ids is None` and `storm_times_unix_sec is None`, will assume that
the input matrices contain probability-matched means.
:param list_of_input_matrices: See doc for
`saliency_maps.read_standard_file`.
:param list_of_saliency_matrices: Same.
:param training_option_dict: Dictionary returned by
`cnn.read_model_metadata`.
:param saliency_colour_map_object: See documentation at top of file.
:param max_colour_value_by_example: length-E numpy array with max value in
colour scheme for each example. Minimum value for [i]th example will be
-1 * max_colour_value_by_example[i], since the colour scheme is
zero-centered and divergent.
:param output_dir_name: Name of output directory (figures will be saved
here).
:param storm_ids: length-E list of storm IDs (strings).
:param storm_times_unix_sec: length-E numpy array of storm times.
"""
pmm_flag = storm_ids is None and storm_times_unix_sec is None
reflectivity_matrix_dbz = list_of_input_matrices[0]
reflectivity_saliency_matrix = list_of_saliency_matrices[0]
az_shear_matrix_s01 = list_of_input_matrices[1]
az_shear_saliency_matrix = list_of_saliency_matrices[1]
num_examples = reflectivity_matrix_dbz.shape[0]
num_reflectivity_heights = len(
training_option_dict[trainval_io.RADAR_HEIGHTS_KEY]
)
num_panel_rows_for_reflectivity = int(numpy.floor(
numpy.sqrt(num_reflectivity_heights)
))
az_shear_field_names = training_option_dict[trainval_io.RADAR_FIELDS_KEY]
num_az_shear_fields = len(az_shear_field_names)
plot_colour_bar_flags = numpy.full(num_az_shear_fields, False, dtype=bool)
for i in range(num_examples):
_, these_axes_objects = radar_plotting.plot_3d_grid_without_coords(
field_matrix=numpy.flip(reflectivity_matrix_dbz[i, ..., 0], axis=0),
field_name=radar_utils.REFL_NAME,
grid_point_heights_metres=training_option_dict[
trainval_io.RADAR_HEIGHTS_KEY],
ground_relative=True,
num_panel_rows=num_panel_rows_for_reflectivity,
font_size=FONT_SIZE_SANS_COLOUR_BARS)
saliency_plotting.plot_many_2d_grids_with_pm_signs(
saliency_matrix_3d=numpy.flip(
reflectivity_saliency_matrix[i, ..., 0], axis=0),
axes_objects_2d_list=these_axes_objects,
colour_map_object=saliency_colour_map_object,
max_absolute_colour_value=max_colour_value_by_example[i])
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(radar_utils.REFL_NAME)
)
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=reflectivity_matrix_dbz[i, ..., 0],
colour_map=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal', extend_min=True, extend_max=True)
if pmm_flag:
this_title_string = 'Probability-matched mean'
this_file_name = '{0:s}/saliency_pmm_reflectivity.jpg'.format(
output_dir_name)
else:
this_storm_time_string = time_conversion.unix_sec_to_string(
storm_times_unix_sec[i], TIME_FORMAT)
this_title_string = 'Storm "{0:s}" at {1:s}'.format(
storm_ids[i], this_storm_time_string)
this_file_name = (
'{0:s}/saliency_{1:s}_{2:s}_reflectivity.jpg'
).format(
output_dir_name, storm_ids[i].replace('_', '-'),
this_storm_time_string
)
this_title_string += ' (max absolute saliency = {0:.3f})'.format(
max_colour_value_by_example[i])
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to file: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
_, these_axes_objects = (
radar_plotting.plot_many_2d_grids_without_coords(
field_matrix=numpy.flip(az_shear_matrix_s01[i, ...], axis=0),
field_name_by_panel=az_shear_field_names,
panel_names=az_shear_field_names, num_panel_rows=1,
plot_colour_bar_by_panel=plot_colour_bar_flags,
font_size=FONT_SIZE_SANS_COLOUR_BARS)
)
saliency_plotting.plot_many_2d_grids_with_pm_signs(
saliency_matrix_3d=numpy.flip(
az_shear_saliency_matrix[i, ...], axis=0),
axes_objects_2d_list=these_axes_objects,
colour_map_object=saliency_colour_map_object,
max_absolute_colour_value=max_colour_value_by_example[i])
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(
radar_utils.LOW_LEVEL_SHEAR_NAME)
)
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=az_shear_saliency_matrix[i, ...],
colour_map=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal', extend_min=True, extend_max=True)
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
this_file_name = this_file_name.replace(
'_reflectivity.jpg', '_azimuthal-shear.jpg')
print 'Saving figure to file: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
def _plot_3d_radar(training_option_dict,
output_dir_name,
pmm_flag,
diff_colour_map_object=None,
max_colour_percentile_for_diff=None,
full_id_strings=None,
storm_time_strings=None,
novel_radar_matrix=None,
novel_radar_matrix_upconv=None,
novel_radar_matrix_upconv_svd=None):
"""Plots results of novelty detection for 3-D radar fields.
E = number of examples (storm objects)
M = number of rows in spatial grid
N = number of columns in spatial grid
H = number of heights in spatial grid
F = number of fields
If `novel_radar_matrix` is the only matrix given, this method will plot the
original (not reconstructed) radar fields.
If `novel_radar_matrix_upconv` is the only matrix given, will plot
upconvnet-reconstructed fields.
If `novel_radar_matrix_upconv_svd` is the only matrix given, will plot
upconvnet-and-SVD-reconstructed fields.
If both `novel_radar_matrix_upconv` and `novel_radar_matrix_upconv_svd` are
given, will plot novelty fields (upconvnet/SVD reconstruction minus
upconvnet reconstruction).
:param training_option_dict: See doc for `cnn.read_model_metadata`.
:param output_dir_name: Name of output directory (figures will be saved
here).
:param pmm_flag: Boolean flag. If True, the input matrices contain
probability-matched means.
:param diff_colour_map_object:
[used only if both `novel_radar_matrix_upconv` and
`novel_radar_matrix_upconv_svd` are given]
See documentation at top of file.
:param max_colour_percentile_for_diff: Same.
:param full_id_strings: [optional and used only if `pmm_flag = False`]
length-E list of full storm IDs.
:param storm_time_strings: [optional and used only if `pmm_flag = False`]
length-E list of storm times.
:param novel_radar_matrix: E-by-M-by-N-by-H-by-F numpy array of original
(not reconstructed) radar fields.
:param novel_radar_matrix_upconv: E-by-M-by-N-by-H-by-F numpy array of
upconvnet-reconstructed radar fields.
:param novel_radar_matrix_upconv_svd: E-by-M-by-N-by-H-by-F numpy array of
upconvnet-and-SVD-reconstructed radar fields.
"""
if pmm_flag:
have_storm_ids = False
else:
have_storm_ids = not (full_id_strings is None
or storm_time_strings is None)
plot_difference = False
if novel_radar_matrix is not None:
plot_type_abbrev = 'actual'
plot_type_verbose = 'actual'
radar_matrix_to_plot = novel_radar_matrix
else:
if (novel_radar_matrix_upconv is not None
and novel_radar_matrix_upconv_svd is not None):
plot_difference = True
plot_type_abbrev = 'novelty'
plot_type_verbose = 'novelty'
radar_matrix_to_plot = (novel_radar_matrix_upconv -
novel_radar_matrix_upconv_svd)
else:
if novel_radar_matrix_upconv is not None:
plot_type_abbrev = 'upconv'
plot_type_verbose = 'upconvnet reconstruction'
radar_matrix_to_plot = novel_radar_matrix_upconv
else:
plot_type_abbrev = 'upconv-svd'
plot_type_verbose = 'upconvnet/SVD reconstruction'
radar_matrix_to_plot = novel_radar_matrix_upconv_svd
radar_field_names = training_option_dict[trainval_io.RADAR_FIELDS_KEY]
radar_heights_m_agl = training_option_dict[trainval_io.RADAR_HEIGHTS_KEY]
num_storms = novel_radar_matrix.shape[0]
num_heights = novel_radar_matrix.shape[-2]
num_panel_rows = int(numpy.floor(numpy.sqrt(num_heights)))
for i in range(num_storms):
if pmm_flag:
this_title_string = 'Probability-matched mean'
this_base_file_name = 'pmm'
else:
if have_storm_ids:
this_title_string = 'Storm "{0:s}" at {1:s}'.format(
full_id_strings[i], storm_time_strings[i])
this_base_file_name = '{0:s}_{1:s}'.format(
full_id_strings[i].replace('_', '-'),
storm_time_strings[i])
else:
this_title_string = 'Example {0:d}'.format(i + 1)
this_base_file_name = 'example{0:06d}'.format(i)
this_title_string += ' ({0:s})'.format(plot_type_verbose)
for j in range(len(radar_field_names)):
this_file_name = '{0:s}/{1:s}_{2:s}_{3:s}.jpg'.format(
output_dir_name, this_base_file_name, plot_type_abbrev,
radar_field_names[j].replace('_', '-'))
if plot_difference:
this_colour_map_object = diff_colour_map_object
this_max_value = numpy.percentile(
numpy.absolute(radar_matrix_to_plot[i, ..., j]),
max_colour_percentile_for_diff)
this_colour_norm_object = matplotlib.colors.Normalize(
vmin=-1 * this_max_value, vmax=this_max_value, clip=False)
else:
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(
radar_field_names[j]))
_, this_axes_object_matrix = (
radar_plotting.plot_3d_grid_without_coords(
field_matrix=numpy.flip(radar_matrix_to_plot[i, ..., j],
axis=0),
field_name=radar_field_names[j],
grid_point_heights_metres=radar_heights_m_agl,
ground_relative=True,
num_panel_rows=num_panel_rows,
font_size=FONT_SIZE_SANS_COLOUR_BARS,
colour_map_object=this_colour_map_object,
colour_norm_object=this_colour_norm_object))
plotting_utils.plot_colour_bar(
axes_object_or_matrix=this_axes_object_matrix,
data_matrix=radar_matrix_to_plot[i, ..., j],
colour_map_object=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation_string='horizontal',
extend_min=True,
extend_max=True)
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print('Saving figure to: "{0:s}"...'.format(this_file_name))
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
def _plot_storm_outlines_one_time(storm_object_table,
valid_time_unix_sec,
axes_object,
basemap_object,
storm_colour,
storm_opacity,
include_secondary_ids,
output_dir_name,
radar_matrix=None,
radar_field_name=None,
radar_latitudes_deg=None,
radar_longitudes_deg=None):
"""Plots storm outlines (and may underlay radar data) at one time step.
M = number of rows in radar grid
N = number of columns in radar grid
K = number of storm objects
:param storm_object_table: See doc for `storm_plotting.plot_storm_outlines`.
:param valid_time_unix_sec: Will plot storm outlines only at this time.
Will plot tracks up to and including this time.
:param axes_object: Same.
:param basemap_object: Same.
:param storm_colour: Same.
:param storm_opacity: Same.
:param include_secondary_ids: Same.
:param output_dir_name: See documentation at top of file.
:param radar_matrix: M-by-N numpy array of radar values. If
`radar_matrix is None`, radar data will simply not be plotted.
:param radar_field_name: [used only if `radar_matrix is not None`]
See documentation at top of file.
:param radar_latitudes_deg: [used only if `radar_matrix is not None`]
length-M numpy array of grid-point latitudes (deg N).
:param radar_longitudes_deg: [used only if `radar_matrix is not None`]
length-N numpy array of grid-point longitudes (deg E).
"""
min_plot_latitude_deg = basemap_object.llcrnrlat
max_plot_latitude_deg = basemap_object.urcrnrlat
min_plot_longitude_deg = basemap_object.llcrnrlon
max_plot_longitude_deg = basemap_object.urcrnrlon
plotting_utils.plot_coastlines(basemap_object=basemap_object,
axes_object=axes_object,
line_colour=BORDER_COLOUR)
plotting_utils.plot_countries(basemap_object=basemap_object,
axes_object=axes_object,
line_colour=BORDER_COLOUR)
plotting_utils.plot_states_and_provinces(basemap_object=basemap_object,
axes_object=axes_object,
line_colour=BORDER_COLOUR)
plotting_utils.plot_parallels(basemap_object=basemap_object,
axes_object=axes_object,
num_parallels=NUM_PARALLELS)
plotting_utils.plot_meridians(basemap_object=basemap_object,
axes_object=axes_object,
num_meridians=NUM_MERIDIANS)
if radar_matrix is not None:
good_indices = numpy.where(
numpy.logical_and(radar_latitudes_deg >= min_plot_latitude_deg,
radar_latitudes_deg <= max_plot_latitude_deg))[0]
radar_latitudes_deg = radar_latitudes_deg[good_indices]
radar_matrix = radar_matrix[good_indices, :]
good_indices = numpy.where(
numpy.logical_and(
radar_longitudes_deg >= min_plot_longitude_deg,
radar_longitudes_deg <= max_plot_longitude_deg))[0]
radar_longitudes_deg = radar_longitudes_deg[good_indices]
radar_matrix = radar_matrix[:, good_indices]
latitude_spacing_deg = radar_latitudes_deg[1] - radar_latitudes_deg[0]
longitude_spacing_deg = (radar_longitudes_deg[1] -
radar_longitudes_deg[0])
radar_plotting.plot_latlng_grid(
field_matrix=radar_matrix,
field_name=radar_field_name,
axes_object=axes_object,
min_grid_point_latitude_deg=numpy.min(radar_latitudes_deg),
min_grid_point_longitude_deg=numpy.min(radar_longitudes_deg),
latitude_spacing_deg=latitude_spacing_deg,
longitude_spacing_deg=longitude_spacing_deg)
colour_map_object, colour_norm_object = (
radar_plotting.get_default_colour_scheme(radar_field_name))
latitude_range_deg = max_plot_latitude_deg - min_plot_latitude_deg
longitude_range_deg = max_plot_longitude_deg - min_plot_longitude_deg
if latitude_range_deg > longitude_range_deg:
orientation_string = 'vertical'
else:
orientation_string = 'horizontal'
colour_bar_object = plotting_utils.plot_colour_bar(
axes_object_or_matrix=axes_object,
data_matrix=radar_matrix,
colour_map_object=colour_map_object,
colour_norm_object=colour_norm_object,
orientation_string=orientation_string,
extend_min=radar_field_name in radar_plotting.SHEAR_VORT_DIV_NAMES,
extend_max=True,
fraction_of_axis_length=0.9)
colour_bar_object.set_label(
radar_plotting.FIELD_NAME_TO_VERBOSE_DICT[radar_field_name])
valid_time_rows = numpy.where(storm_object_table[
tracking_utils.VALID_TIME_COLUMN].values == valid_time_unix_sec)[0]
line_colour = matplotlib.colors.to_rgba(storm_colour, storm_opacity)
storm_plotting.plot_storm_outlines(
storm_object_table=storm_object_table.iloc[valid_time_rows],
axes_object=axes_object,
basemap_object=basemap_object,
line_colour=line_colour)
storm_plotting.plot_storm_ids(
storm_object_table=storm_object_table.iloc[valid_time_rows],
axes_object=axes_object,
basemap_object=basemap_object,
plot_near_centroids=False,
include_secondary_ids=include_secondary_ids,
font_colour=storm_plotting.DEFAULT_FONT_COLOUR)
storm_plotting.plot_storm_tracks(storm_object_table=storm_object_table,
axes_object=axes_object,
basemap_object=basemap_object,
colour_map_object=None,
line_colour=TRACK_COLOUR)
nice_time_string = time_conversion.unix_sec_to_string(
valid_time_unix_sec, NICE_TIME_FORMAT)
abbrev_time_string = time_conversion.unix_sec_to_string(
valid_time_unix_sec, FILE_NAME_TIME_FORMAT)
pyplot.title('Storm objects at {0:s}'.format(nice_time_string))
output_file_name = '{0:s}/storm_outlines_{1:s}.jpg'.format(
output_dir_name, abbrev_time_string)
print('Saving figure to: "{0:s}"...'.format(output_file_name))
pyplot.savefig(output_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
imagemagick_utils.trim_whitespace(input_file_name=output_file_name,
output_file_name=output_file_name)
def _plot_saliency_for_3d_radar(
radar_matrix, radar_saliency_matrix, model_metadata_dict,
saliency_colour_map_object, max_colour_value_by_example,
output_dir_name, storm_ids=None, storm_times_unix_sec=None):
"""Plots saliency for 3-D radar fields.
E = number of examples
M = number of rows in spatial grid
N = number of columns in spatial grid
H = number of heights in spatial grid
F = number of fields
If `storm_ids is None` and `storm_times_unix_sec is None`, will assume that
the input matrices contain probability-matched means.
:param radar_matrix: E-by-M-by-N-by-H-by-F numpy array of radar values
(predictors).
:param radar_saliency_matrix: E-by-M-by-N-by-H-by-F numpy array of saliency
values.
:param model_metadata_dict: See doc for `cnn.read_model_metadata`.
:param saliency_colour_map_object: See doc for
`_plot_saliency_for_2d3d_radar`.
:param max_colour_value_by_example: Same.
:param output_dir_name: Same.
:param storm_ids: Same.
:param storm_times_unix_sec: Same.
"""
pmm_flag = storm_ids is None and storm_times_unix_sec is None
training_option_dict = model_metadata_dict[cnn.TRAINING_OPTION_DICT_KEY]
num_examples = radar_matrix.shape[0]
num_fields = radar_matrix.shape[-1]
num_heights = len(training_option_dict[trainval_io.RADAR_HEIGHTS_KEY])
num_panel_rows = int(numpy.floor(numpy.sqrt(num_heights)))
for i in range(num_examples):
for k in range(num_fields):
this_field_name = training_option_dict[
trainval_io.RADAR_FIELDS_KEY][k]
_, these_axes_objects = (
radar_plotting.plot_3d_grid_without_coords(
field_matrix=numpy.flip(radar_matrix[i, ..., k], axis=0),
field_name=this_field_name,
grid_point_heights_metres=training_option_dict[
trainval_io.RADAR_HEIGHTS_KEY],
ground_relative=True, num_panel_rows=num_panel_rows,
font_size=FONT_SIZE_SANS_COLOUR_BARS)
)
saliency_plotting.plot_many_2d_grids_with_pm_signs(
saliency_matrix_3d=numpy.flip(
radar_saliency_matrix[i, ..., k], axis=0),
axes_objects_2d_list=these_axes_objects,
colour_map_object=saliency_colour_map_object,
max_absolute_colour_value=max_colour_value_by_example[i])
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(this_field_name)
)
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=radar_matrix[i, ..., k],
colour_map=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal', extend_min=True, extend_max=True)
if pmm_flag:
this_title_string = 'Probability-matched mean'
this_file_name = '{0:s}/saliency_pmm_{1:s}.jpg'.format(
output_dir_name, this_field_name.replace('_', '-')
)
else:
this_storm_time_string = time_conversion.unix_sec_to_string(
storm_times_unix_sec[i], TIME_FORMAT)
this_title_string = 'Storm "{0:s}" at {1:s}'.format(
storm_ids[i], this_storm_time_string)
this_file_name = '{0:s}/saliency_{1:s}_{2:s}_{3:s}.jpg'.format(
output_dir_name, storm_ids[i].replace('_', '-'),
this_storm_time_string, this_field_name.replace('_', '-')
)
this_title_string += ' (max absolute saliency = {0:.3f})'.format(
max_colour_value_by_example[i])
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to file: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
def _plot_3d_radar_cams(
radar_matrix, model_metadata_dict, cam_colour_map_object,
max_colour_prctile_for_cam, output_dir_name,
class_activation_matrix=None, ggradcam_output_matrix=None,
storm_ids=None, storm_times_unix_sec=None):
"""Plots class-activation maps for 3-D radar data.
E = number of examples
M = number of rows in spatial grid
N = number of columns in spatial grid
H = number of heights in spatial grid
F = number of radar fields
This method will plot either `class_activation_matrix` or
`ggradcam_output_matrix`, not both.
If `storm_ids is None` and `storm_times_unix_sec is None`, will assume that
the input matrices contain probability-matched means.
:param radar_matrix: E-by-M-by-N-by-H-by-F numpy array of radar values.
:param model_metadata_dict: Dictionary with CNN metadata (see doc for
`cnn.read_model_metadata`).
:param cam_colour_map_object: See documentation at top of file.
:param max_colour_prctile_for_cam: Same.
:param output_dir_name: Same.
:param class_activation_matrix: E-by-M-by-N-by-H numpy array of class
activations.
:param ggradcam_output_matrix: E-by-M-by-N-by-H-by-F numpy array of output
values from guided Grad-CAM.
:param storm_ids: length-E list of storm IDs (strings).
:param storm_times_unix_sec: length-E numpy array of storm times.
"""
pmm_flag = storm_ids is None and storm_times_unix_sec is None
num_examples = radar_matrix.shape[0]
num_heights = radar_matrix.shape[-2]
num_fields = radar_matrix.shape[-1]
num_panel_rows = int(numpy.floor(numpy.sqrt(num_heights)))
if class_activation_matrix is None:
quantity_string = 'max absolute guided Grad-CAM output'
pathless_file_name_prefix = 'guided-gradcam'
else:
quantity_string = 'max class activation'
pathless_file_name_prefix = 'gradcam'
training_option_dict = model_metadata_dict[cnn.TRAINING_OPTION_DICT_KEY]
for i in range(num_examples):
for k in range(num_fields):
this_field_name = training_option_dict[
trainval_io.RADAR_FIELDS_KEY][k]
_, these_axes_objects = radar_plotting.plot_3d_grid_without_coords(
field_matrix=numpy.flip(radar_matrix[i, ..., k], axis=0),
field_name=this_field_name,
grid_point_heights_metres=training_option_dict[
trainval_io.RADAR_HEIGHTS_KEY],
ground_relative=True, num_panel_rows=num_panel_rows,
font_size=FONT_SIZE_SANS_COLOUR_BARS)
if class_activation_matrix is None:
this_matrix = ggradcam_output_matrix[i, ..., k]
this_max_contour_level = numpy.percentile(
numpy.absolute(this_matrix), max_colour_prctile_for_cam)
if this_max_contour_level == 0:
this_max_contour_level = 10.
saliency_plotting.plot_many_2d_grids_with_contours(
saliency_matrix_3d=numpy.flip(this_matrix, axis=0),
axes_objects_2d_list=these_axes_objects,
colour_map_object=cam_colour_map_object,
max_absolute_contour_level=this_max_contour_level,
contour_interval=this_max_contour_level / 10)
else:
this_matrix = class_activation_matrix[i, ...]
this_max_contour_level = numpy.percentile(
this_matrix, max_colour_prctile_for_cam)
if this_max_contour_level == 0:
this_max_contour_level = 10.
cam_plotting.plot_many_2d_grids(
class_activation_matrix_3d=numpy.flip(this_matrix, axis=0),
axes_objects_2d_list=these_axes_objects,
colour_map_object=cam_colour_map_object,
max_contour_level=this_max_contour_level,
contour_interval=this_max_contour_level / NUM_CONTOURS)
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(this_field_name)
)
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=radar_matrix[i, ..., k],
colour_map=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal', extend_min=True, extend_max=True)
if pmm_flag:
this_title_string = 'Probability-matched mean'
this_figure_file_name = '{0:s}/{1:s}_pmm_{2:s}.jpg'.format(
output_dir_name, pathless_file_name_prefix,
this_field_name.replace('_', '-')
)
else:
this_storm_time_string = time_conversion.unix_sec_to_string(
storm_times_unix_sec[i], TIME_FORMAT)
this_title_string = 'Storm "{0:s}" at {1:s}'.format(
storm_ids[i], this_storm_time_string)
this_figure_file_name = (
'{0:s}/{1:s}_{2:s}_{3:s}_{4:s}.jpg'
).format(
output_dir_name, pathless_file_name_prefix,
storm_ids[i].replace('_', '-'), this_storm_time_string,
this_field_name.replace('_', '-')
)
this_title_string += ' ({0:s} = {1:.3f})'.format(
quantity_string, this_max_contour_level)
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to file: "{0:s}"...'.format(
this_figure_file_name)
pyplot.savefig(this_figure_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
def _plot_storm_outlines_one_time(storm_object_table,
valid_time_unix_sec,
warning_table,
axes_object,
basemap_object,
storm_outline_colour,
storm_outline_opacity,
include_secondary_ids,
output_dir_name,
primary_id_to_track_colour=None,
radar_matrix=None,
radar_field_name=None,
radar_latitudes_deg=None,
radar_longitudes_deg=None,
radar_colour_map_object=None):
"""Plots storm outlines (and may underlay radar data) at one time step.
M = number of rows in radar grid
N = number of columns in radar grid
K = number of storm objects
If `primary_id_to_track_colour is None`, all storm tracks will be the same
colour.
:param storm_object_table: See doc for `storm_plotting.plot_storm_outlines`.
:param valid_time_unix_sec: Will plot storm outlines only at this time.
Will plot tracks up to and including this time.
:param warning_table: None or a pandas table with the following columns.
warning_table.start_time_unix_sec: Start time.
warning_table.end_time_unix_sec: End time.
warning_table.polygon_object_latlng: Polygon (instance of
`shapely.geometry.Polygon`) with lat-long coordinates of warning
boundary.
:param axes_object: See doc for `storm_plotting.plot_storm_outlines`.
:param basemap_object: Same.
:param storm_outline_colour: Same.
:param storm_outline_opacity: Same.
:param include_secondary_ids: Same.
:param output_dir_name: See documentation at top of file.
:param primary_id_to_track_colour: Dictionary created by
`_assign_colours_to_storms`. If this is None, all storm tracks will be
the same colour.
:param radar_matrix: M-by-N numpy array of radar values. If
`radar_matrix is None`, radar data will simply not be plotted.
:param radar_field_name: [used only if `radar_matrix is not None`]
See documentation at top of file.
:param radar_latitudes_deg: [used only if `radar_matrix is not None`]
length-M numpy array of grid-point latitudes (deg N).
:param radar_longitudes_deg: [used only if `radar_matrix is not None`]
length-N numpy array of grid-point longitudes (deg E).
:param radar_colour_map_object: [used only if `radar_matrix is not None`]
Colour map (instance of `matplotlib.pyplot.cm`). If None, will use
default for the given field.
"""
# plot_storm_ids = radar_matrix is None or radar_colour_map_object is None
plot_storm_ids = False
min_plot_latitude_deg = basemap_object.llcrnrlat
max_plot_latitude_deg = basemap_object.urcrnrlat
min_plot_longitude_deg = basemap_object.llcrnrlon
max_plot_longitude_deg = basemap_object.urcrnrlon
plotting_utils.plot_coastlines(basemap_object=basemap_object,
axes_object=axes_object,
line_colour=BORDER_COLOUR)
plotting_utils.plot_countries(basemap_object=basemap_object,
axes_object=axes_object,
line_colour=BORDER_COLOUR)
plotting_utils.plot_states_and_provinces(basemap_object=basemap_object,
axes_object=axes_object,
line_colour=BORDER_COLOUR)
plotting_utils.plot_parallels(basemap_object=basemap_object,
axes_object=axes_object,
num_parallels=NUM_PARALLELS)
plotting_utils.plot_meridians(basemap_object=basemap_object,
axes_object=axes_object,
num_meridians=NUM_MERIDIANS)
if radar_matrix is not None:
custom_colour_map = radar_colour_map_object is not None
good_indices = numpy.where(
numpy.logical_and(radar_latitudes_deg >= min_plot_latitude_deg,
radar_latitudes_deg <= max_plot_latitude_deg))[0]
radar_latitudes_deg = radar_latitudes_deg[good_indices]
radar_matrix = radar_matrix[good_indices, :]
good_indices = numpy.where(
numpy.logical_and(
radar_longitudes_deg >= min_plot_longitude_deg,
radar_longitudes_deg <= max_plot_longitude_deg))[0]
radar_longitudes_deg = radar_longitudes_deg[good_indices]
radar_matrix = radar_matrix[:, good_indices]
latitude_spacing_deg = radar_latitudes_deg[1] - radar_latitudes_deg[0]
longitude_spacing_deg = (radar_longitudes_deg[1] -
radar_longitudes_deg[0])
if radar_colour_map_object is None:
colour_map_object, colour_norm_object = (
radar_plotting.get_default_colour_scheme(radar_field_name))
else:
colour_map_object = radar_colour_map_object
colour_norm_object = radar_plotting.get_default_colour_scheme(
radar_field_name)[-1]
this_ratio = radar_plotting._field_to_plotting_units(
field_matrix=1., field_name=radar_field_name)
colour_norm_object = pyplot.Normalize(
colour_norm_object.vmin / this_ratio,
colour_norm_object.vmax / this_ratio)
radar_plotting.plot_latlng_grid(
field_matrix=radar_matrix,
field_name=radar_field_name,
axes_object=axes_object,
min_grid_point_latitude_deg=numpy.min(radar_latitudes_deg),
min_grid_point_longitude_deg=numpy.min(radar_longitudes_deg),
latitude_spacing_deg=latitude_spacing_deg,
longitude_spacing_deg=longitude_spacing_deg,
colour_map_object=colour_map_object,
colour_norm_object=colour_norm_object)
latitude_range_deg = max_plot_latitude_deg - min_plot_latitude_deg
longitude_range_deg = max_plot_longitude_deg - min_plot_longitude_deg
if latitude_range_deg > longitude_range_deg:
orientation_string = 'vertical'
else:
orientation_string = 'horizontal'
colour_bar_object = plotting_utils.plot_colour_bar(
axes_object_or_matrix=axes_object,
data_matrix=radar_matrix,
colour_map_object=colour_map_object,
colour_norm_object=colour_norm_object,
orientation_string=orientation_string,
padding=0.05,
extend_min=radar_field_name in radar_plotting.SHEAR_VORT_DIV_NAMES,
extend_max=True,
fraction_of_axis_length=1.)
radar_field_name_verbose = radar_utils.field_name_to_verbose(
field_name=radar_field_name, include_units=True)
radar_field_name_verbose = radar_field_name_verbose.replace(
'm ASL', 'kft ASL')
colour_bar_object.set_label(radar_field_name_verbose)
if custom_colour_map:
tick_values = colour_bar_object.get_ticks()
tick_label_strings = ['{0:.1f}'.format(v) for v in tick_values]
colour_bar_object.set_ticks(tick_values)
colour_bar_object.set_ticklabels(tick_label_strings)
valid_time_rows = numpy.where(storm_object_table[
tracking_utils.VALID_TIME_COLUMN].values == valid_time_unix_sec)[0]
this_colour = matplotlib.colors.to_rgba(storm_outline_colour,
storm_outline_opacity)
storm_plotting.plot_storm_outlines(
storm_object_table=storm_object_table.iloc[valid_time_rows],
axes_object=axes_object,
basemap_object=basemap_object,
line_colour=this_colour)
if plot_storm_ids:
storm_plotting.plot_storm_ids(
storm_object_table=storm_object_table.iloc[valid_time_rows],
axes_object=axes_object,
basemap_object=basemap_object,
plot_near_centroids=False,
include_secondary_ids=include_secondary_ids,
font_colour=storm_plotting.DEFAULT_FONT_COLOUR)
if warning_table is not None:
warning_indices = numpy.where(
numpy.logical_and(
warning_table[WARNING_START_TIME_KEY].values <=
valid_time_unix_sec, warning_table[WARNING_END_TIME_KEY].values
>= valid_time_unix_sec))[0]
for k in warning_indices:
this_vertex_dict = polygons.polygon_object_to_vertex_arrays(
warning_table[WARNING_LATLNG_POLYGON_KEY].values[k])
these_latitudes_deg = this_vertex_dict[polygons.EXTERIOR_Y_COLUMN]
these_longitudes_deg = this_vertex_dict[polygons.EXTERIOR_X_COLUMN]
these_latitude_flags = numpy.logical_and(
these_latitudes_deg >= min_plot_latitude_deg,
these_latitudes_deg <= max_plot_latitude_deg)
these_longitude_flags = numpy.logical_and(
these_longitudes_deg >= min_plot_longitude_deg,
these_longitudes_deg <= max_plot_longitude_deg)
these_coord_flags = numpy.logical_and(these_latitude_flags,
these_longitude_flags)
if not numpy.any(these_coord_flags):
continue
these_x_metres, these_y_metres = basemap_object(
these_longitudes_deg, these_latitudes_deg)
axes_object.plot(these_x_metres,
these_y_metres,
color=this_colour,
linestyle='dashed',
linewidth=storm_plotting.DEFAULT_POLYGON_WIDTH)
axes_object.text(numpy.mean(these_x_metres),
numpy.mean(these_y_metres),
'W{0:d}'.format(k),
fontsize=storm_plotting.DEFAULT_FONT_SIZE,
fontweight='bold',
color=this_colour,
horizontalalignment='center',
verticalalignment='center')
these_sec_id_strings = (
warning_table[LINKED_SECONDARY_IDS_KEY].values[k])
if len(these_sec_id_strings) == 0:
continue
these_object_indices = numpy.array([], dtype=int)
for this_sec_id_string in these_sec_id_strings:
these_subindices = numpy.where(
storm_object_table[tracking_utils.SECONDARY_ID_COLUMN].
values[valid_time_rows] == this_sec_id_string)[0]
these_object_indices = numpy.concatenate(
(these_object_indices, valid_time_rows[these_subindices]))
for i in these_object_indices:
this_vertex_dict = polygons.polygon_object_to_vertex_arrays(
storm_object_table[
tracking_utils.LATLNG_POLYGON_COLUMN].values[i])
these_x_metres, these_y_metres = basemap_object(
this_vertex_dict[polygons.EXTERIOR_X_COLUMN],
this_vertex_dict[polygons.EXTERIOR_Y_COLUMN])
axes_object.text(numpy.mean(these_x_metres),
numpy.mean(these_y_metres),
'W{0:d}'.format(k),
fontsize=storm_plotting.DEFAULT_FONT_SIZE,
fontweight='bold',
color=this_colour,
horizontalalignment='center',
verticalalignment='center')
if primary_id_to_track_colour is None:
storm_plotting.plot_storm_tracks(storm_object_table=storm_object_table,
axes_object=axes_object,
basemap_object=basemap_object,
colour_map_object=None,
constant_colour=DEFAULT_TRACK_COLOUR)
else:
for this_primary_id_string in primary_id_to_track_colour:
this_storm_object_table = storm_object_table.loc[
storm_object_table[tracking_utils.PRIMARY_ID_COLUMN] ==
this_primary_id_string]
if len(this_storm_object_table.index) == 0:
continue
storm_plotting.plot_storm_tracks(
storm_object_table=this_storm_object_table,
axes_object=axes_object,
basemap_object=basemap_object,
colour_map_object=None,
constant_colour=primary_id_to_track_colour[
this_primary_id_string])
nice_time_string = time_conversion.unix_sec_to_string(
valid_time_unix_sec, NICE_TIME_FORMAT)
abbrev_time_string = time_conversion.unix_sec_to_string(
valid_time_unix_sec, FILE_NAME_TIME_FORMAT)
pyplot.title('Storm objects at {0:s}'.format(nice_time_string))
output_file_name = '{0:s}/storm_outlines_{1:s}.jpg'.format(
output_dir_name, abbrev_time_string)
print('Saving figure to: "{0:s}"...'.format(output_file_name))
pyplot.savefig(output_file_name,
dpi=FIGURE_RESOLUTION_DPI,
pad_inches=0,
bbox_inches='tight')
pyplot.close()
def _plot_bwo_for_3d_radar(
optimized_radar_matrix, training_option_dict, diff_colour_map_object,
max_colour_percentile_for_diff, top_output_dir_name, pmm_flag,
input_radar_matrix=None, storm_ids=None, storm_times_unix_sec=None):
"""Plots BWO results for 3-D radar fields.
E = number of examples (storm objects)
M = number of rows in spatial grid
N = number of columns in spatial grid
H = number of heights in spatial grid
F = number of fields
:param optimized_radar_matrix: E-by-M-by-N-by-H-by-F numpy array of radar
values (predictors).
:param training_option_dict: See doc for `_plot_bwo_for_2d3d_radar`.
:param diff_colour_map_object: Same.
:param max_colour_percentile_for_diff: Same.
:param top_output_dir_name: Same.
:param pmm_flag: Same.
:param input_radar_matrix: Same as `optimized_radar_matrix` but with
non-optimized input.
:param storm_ids: See doc for `_plot_bwo_for_2d3d_radar`.
:param storm_times_unix_sec: Same.
"""
before_optimization_dir_name = '{0:s}/before_optimization'.format(
top_output_dir_name)
after_optimization_dir_name = '{0:s}/after_optimization'.format(
top_output_dir_name)
difference_dir_name = '{0:s}/after_minus_before_optimization'.format(
top_output_dir_name)
file_system_utils.mkdir_recursive_if_necessary(
directory_name=before_optimization_dir_name)
file_system_utils.mkdir_recursive_if_necessary(
directory_name=after_optimization_dir_name)
file_system_utils.mkdir_recursive_if_necessary(
directory_name=difference_dir_name)
if pmm_flag:
have_storm_ids = False
else:
have_storm_ids = not (storm_ids is None or storm_times_unix_sec is None)
radar_field_names = training_option_dict[trainval_io.RADAR_FIELDS_KEY]
radar_heights_m_agl = training_option_dict[trainval_io.RADAR_HEIGHTS_KEY]
num_storms = optimized_radar_matrix.shape[0]
num_heights = optimized_radar_matrix.shape[-2]
num_panel_rows = int(numpy.floor(
numpy.sqrt(num_heights)
))
for i in range(num_storms):
print '\n'
if pmm_flag:
this_base_title_string = 'Probability-matched mean'
this_base_pathless_file_name = 'pmm'
else:
if have_storm_ids:
this_storm_time_string = time_conversion.unix_sec_to_string(
storm_times_unix_sec[i], TIME_FORMAT)
this_base_title_string = 'Storm "{0:s}" at {1:s}'.format(
storm_ids[i], this_storm_time_string)
this_base_pathless_file_name = '{0:s}_{1:s}'.format(
storm_ids[i].replace('_', '-'), this_storm_time_string)
else:
this_base_title_string = 'Example {0:d}'.format(i + 1)
this_base_pathless_file_name = 'example{0:06d}'.format(i)
for j in range(len(radar_field_names)):
_, these_axes_objects = (
radar_plotting.plot_3d_grid_without_coords(
field_matrix=numpy.flip(
optimized_radar_matrix[i, ..., j], axis=0),
field_name=radar_field_names[j],
grid_point_heights_metres=radar_heights_m_agl,
ground_relative=True, num_panel_rows=num_panel_rows,
font_size=FONT_SIZE_SANS_COLOUR_BARS)
)
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(
radar_field_names[j])
)
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=optimized_radar_matrix[i, ..., j],
colour_map=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal', extend_min=True, extend_max=True)
this_title_string = '{0:s} (after optimization)'.format(
this_base_title_string)
this_file_name = (
'{0:s}/{1:s}_after-optimization_{2:s}.jpg'
).format(
after_optimization_dir_name, this_base_pathless_file_name,
radar_field_names[j].replace('_', '-')
)
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
if input_radar_matrix is None:
continue
_, these_axes_objects = (
radar_plotting.plot_3d_grid_without_coords(
field_matrix=numpy.flip(
input_radar_matrix[i, ..., j], axis=0),
field_name=radar_field_names[j],
grid_point_heights_metres=radar_heights_m_agl,
ground_relative=True, num_panel_rows=num_panel_rows,
font_size=FONT_SIZE_SANS_COLOUR_BARS)
)
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(
radar_field_names[j])
)
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=input_radar_matrix[i, ..., j],
colour_map=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal', extend_min=True, extend_max=True)
this_title_string = '{0:s} (before optimization)'.format(
this_base_title_string)
this_file_name = (
'{0:s}/{1:s}_before-optimization_{2:s}.jpg'
).format(
before_optimization_dir_name, this_base_pathless_file_name,
radar_field_names[j].replace('_', '-')
)
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
this_diff_matrix = (
optimized_radar_matrix[i, ..., j] -
input_radar_matrix[i, ..., j]
)
this_max_value = numpy.percentile(
numpy.absolute(this_diff_matrix),
max_colour_percentile_for_diff)
this_colour_norm_object = matplotlib.colors.Normalize(
vmin=-1 * this_max_value, vmax=this_max_value, clip=False)
_, these_axes_objects = (
radar_plotting.plot_3d_grid_without_coords(
field_matrix=numpy.flip(this_diff_matrix, axis=0),
field_name=radar_field_names[j],
grid_point_heights_metres=radar_heights_m_agl,
ground_relative=True, num_panel_rows=num_panel_rows,
font_size=FONT_SIZE_SANS_COLOUR_BARS,
colour_map_object=diff_colour_map_object,
colour_norm_object=this_colour_norm_object)
)
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=this_diff_matrix,
colour_map=diff_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal', extend_min=True, extend_max=True)
this_title_string = (
'{0:s} (after minus before optimization)'
).format(this_base_title_string)
this_file_name = (
'{0:s}/{1:s}_optimization-diff_{2:s}.jpg'
).format(
difference_dir_name, this_base_pathless_file_name,
radar_field_names[j].replace('_', '-')
)
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
def _plot_bwo_for_2d3d_radar(
list_of_optimized_matrices, training_option_dict,
diff_colour_map_object, max_colour_percentile_for_diff,
top_output_dir_name, pmm_flag, list_of_input_matrices=None,
storm_ids=None, storm_times_unix_sec=None):
"""Plots BWO results for 2-D azimuthal-shear and 3-D reflectivity fields.
E = number of examples (storm objects)
T = number of input tensors to the model
:param list_of_optimized_matrices: length-T list of numpy arrays, where the
[i]th array is the optimized version of the [i]th input matrix to the
model.
:param training_option_dict: See doc for `cnn.read_model_metadata`.
:param diff_colour_map_object: See documentation at top of file.
:param max_colour_percentile_for_diff: Same.
:param top_output_dir_name: Path to top-level output directory (figures will
be saved here).
:param pmm_flag: Boolean flag. If True, `list_of_predictor_matrices`
contains probability-matched means.
:param list_of_input_matrices: Same as `list_of_optimized_matrices` but with
non-optimized input matrices.
:param storm_ids: [optional and used only if `pmm_flag = False`]
length-E list of storm IDs (strings).
:param storm_times_unix_sec: [optional and used only if `pmm_flag = False`]
length-E numpy array of storm times.
"""
before_optimization_dir_name = '{0:s}/before_optimization'.format(
top_output_dir_name)
after_optimization_dir_name = '{0:s}/after_optimization'.format(
top_output_dir_name)
difference_dir_name = '{0:s}/after_minus_before_optimization'.format(
top_output_dir_name)
file_system_utils.mkdir_recursive_if_necessary(
directory_name=before_optimization_dir_name)
file_system_utils.mkdir_recursive_if_necessary(
directory_name=after_optimization_dir_name)
file_system_utils.mkdir_recursive_if_necessary(
directory_name=difference_dir_name)
if pmm_flag:
have_storm_ids = False
else:
have_storm_ids = not (storm_ids is None or storm_times_unix_sec is None)
az_shear_field_names = training_option_dict[trainval_io.RADAR_FIELDS_KEY]
num_az_shear_fields = len(az_shear_field_names)
plot_colour_bar_flags = numpy.full(num_az_shear_fields, False, dtype=bool)
num_storms = list_of_optimized_matrices[0].shape[0]
for i in range(num_storms):
print '\n'
if pmm_flag:
this_base_title_string = 'Probability-matched mean'
this_base_pathless_file_name = 'pmm'
else:
if have_storm_ids:
this_storm_time_string = time_conversion.unix_sec_to_string(
storm_times_unix_sec[i], TIME_FORMAT)
this_base_title_string = 'Storm "{0:s}" at {1:s}'.format(
storm_ids[i], this_storm_time_string)
this_base_pathless_file_name = '{0:s}_{1:s}'.format(
storm_ids[i].replace('_', '-'), this_storm_time_string)
else:
this_base_title_string = 'Example {0:d}'.format(i + 1)
this_base_pathless_file_name = 'example{0:06d}'.format(i)
this_reflectivity_matrix_dbz = numpy.flip(
list_of_optimized_matrices[0][i, ..., 0], axis=0)
this_num_heights = this_reflectivity_matrix_dbz.shape[-1]
this_num_panel_rows = int(numpy.floor(
numpy.sqrt(this_num_heights)
))
_, these_axes_objects = radar_plotting.plot_3d_grid_without_coords(
field_matrix=this_reflectivity_matrix_dbz,
field_name=radar_utils.REFL_NAME,
grid_point_heights_metres=training_option_dict[
trainval_io.RADAR_HEIGHTS_KEY],
ground_relative=True, num_panel_rows=this_num_panel_rows,
font_size=FONT_SIZE_SANS_COLOUR_BARS)
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(radar_utils.REFL_NAME)
)
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=this_reflectivity_matrix_dbz,
colour_map=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal', extend_min=True, extend_max=True)
this_title_string = '{0:s} (after optimization)'.format(
this_base_title_string)
this_file_name = (
'{0:s}/{1:s}_after-optimization_reflectivity.jpg'
).format(after_optimization_dir_name, this_base_pathless_file_name)
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
this_az_shear_matrix_s01 = numpy.flip(
list_of_optimized_matrices[1][i, ..., 0], axis=0)
_, these_axes_objects = (
radar_plotting.plot_many_2d_grids_without_coords(
field_matrix=this_az_shear_matrix_s01,
field_name_by_panel=az_shear_field_names, num_panel_rows=1,
panel_names=az_shear_field_names,
plot_colour_bar_by_panel=plot_colour_bar_flags,
font_size=FONT_SIZE_SANS_COLOUR_BARS)
)
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(
radar_utils.LOW_LEVEL_SHEAR_NAME)
)
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=this_az_shear_matrix_s01,
colour_map=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal', extend_min=True, extend_max=True)
this_title_string = '{0:s} (after optimization)'.format(
this_base_title_string)
this_file_name = (
'{0:s}/{1:s}_after-optimization_azimuthal-shear.jpg'
).format(after_optimization_dir_name, this_base_pathless_file_name)
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
if list_of_input_matrices is None:
continue
this_reflectivity_matrix_dbz = numpy.flip(
list_of_input_matrices[0][i, ..., 0], axis=0)
_, these_axes_objects = radar_plotting.plot_3d_grid_without_coords(
field_matrix=this_reflectivity_matrix_dbz,
field_name=radar_utils.REFL_NAME,
grid_point_heights_metres=training_option_dict[
trainval_io.RADAR_HEIGHTS_KEY],
ground_relative=True, num_panel_rows=this_num_panel_rows,
font_size=FONT_SIZE_SANS_COLOUR_BARS)
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(radar_utils.REFL_NAME)
)
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=this_reflectivity_matrix_dbz,
colour_map=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal', extend_min=True, extend_max=True)
this_title_string = '{0:s} (before optimization)'.format(
this_base_title_string)
this_file_name = (
'{0:s}/{1:s}_before-optimization_reflectivity.jpg'
).format(before_optimization_dir_name, this_base_pathless_file_name)
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
this_az_shear_matrix_s01 = numpy.flip(
list_of_input_matrices[1][i, ..., 0], axis=0)
_, these_axes_objects = (
radar_plotting.plot_many_2d_grids_without_coords(
field_matrix=this_az_shear_matrix_s01,
field_name_by_panel=az_shear_field_names, num_panel_rows=1,
panel_names=az_shear_field_names,
plot_colour_bar_by_panel=plot_colour_bar_flags,
font_size=FONT_SIZE_SANS_COLOUR_BARS)
)
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(
radar_utils.LOW_LEVEL_SHEAR_NAME)
)
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=this_az_shear_matrix_s01,
colour_map=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal', extend_min=True, extend_max=True)
this_title_string = '{0:s} (before optimization)'.format(
this_base_title_string)
this_file_name = (
'{0:s}/{1:s}_before-optimization_azimuthal-shear.jpg'
).format(before_optimization_dir_name, this_base_pathless_file_name)
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
this_refl_diff_matrix_dbz = (
list_of_optimized_matrices[0][i, ..., 0] -
list_of_input_matrices[0][i, ..., 0]
)
this_refl_diff_matrix_dbz = numpy.flip(
this_refl_diff_matrix_dbz, axis=0)
this_max_value_dbz = numpy.percentile(
numpy.absolute(this_refl_diff_matrix_dbz),
max_colour_percentile_for_diff)
this_colour_norm_object = matplotlib.colors.Normalize(
vmin=-1 * this_max_value_dbz, vmax=this_max_value_dbz, clip=False)
_, these_axes_objects = radar_plotting.plot_3d_grid_without_coords(
field_matrix=this_refl_diff_matrix_dbz,
field_name=radar_utils.REFL_NAME,
grid_point_heights_metres=training_option_dict[
trainval_io.RADAR_HEIGHTS_KEY],
ground_relative=True, num_panel_rows=this_num_panel_rows,
font_size=FONT_SIZE_SANS_COLOUR_BARS,
colour_map_object=diff_colour_map_object,
colour_norm_object=this_colour_norm_object)
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=this_refl_diff_matrix_dbz,
colour_map=diff_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal', extend_min=True, extend_max=True)
this_title_string = '{0:s} (after minus before optimization)'.format(
this_base_title_string)
this_file_name = (
'{0:s}/{1:s}_optimization-diff_reflectivity.jpg'
).format(difference_dir_name, this_base_pathless_file_name)
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
this_shear_diff_matrix_s01 = (
list_of_optimized_matrices[1][i, ..., 0] -
list_of_input_matrices[1][i, ..., 0]
)
this_shear_diff_matrix_s01 = numpy.flip(
this_shear_diff_matrix_s01, axis=0)
this_max_value_s01 = numpy.percentile(
numpy.absolute(this_shear_diff_matrix_s01),
max_colour_percentile_for_diff)
this_colour_norm_object = matplotlib.colors.Normalize(
vmin=-1 * this_max_value_s01, vmax=this_max_value_s01, clip=False)
_, these_axes_objects = (
radar_plotting.plot_many_2d_grids_without_coords(
field_matrix=this_shear_diff_matrix_s01,
field_name_by_panel=az_shear_field_names, num_panel_rows=1,
panel_names=az_shear_field_names,
colour_map_object_by_panel=
[diff_colour_map_object] * num_az_shear_fields,
colour_norm_object_by_panel=
[copy.deepcopy(this_colour_norm_object)] * num_az_shear_fields,
plot_colour_bar_by_panel=plot_colour_bar_flags,
font_size=FONT_SIZE_SANS_COLOUR_BARS)
)
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=this_shear_diff_matrix_s01,
colour_map=diff_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal', extend_min=True, extend_max=True)
this_title_string = '{0:s} (after minus before optimization)'.format(
this_base_title_string)
this_file_name = (
'{0:s}/{1:s}_optimization-diff_azimuthal-shear.jpg'
).format(difference_dir_name, this_base_pathless_file_name)
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
def _plot_one_field(reflectivity_matrix_dbz, latitudes_deg, longitudes_deg,
add_colour_bar, panel_letter, output_file_name):
"""Plots reflectivity field from one dataset.
:param reflectivity_matrix_dbz: See doc for `_read_file`.
:param latitudes_deg: Same.
:param longitudes_deg: Same.
:param add_colour_bar: Boolean flag.
:param panel_letter: Panel letter (will be printed at top left of figure).
:param output_file_name: Path to output file (figure will be saved here).
"""
(figure_object, axes_object,
basemap_object) = plotting_utils.create_equidist_cylindrical_map(
min_latitude_deg=numpy.min(latitudes_deg),
max_latitude_deg=numpy.max(latitudes_deg),
min_longitude_deg=numpy.min(longitudes_deg),
max_longitude_deg=numpy.max(longitudes_deg),
resolution_string='i')
plotting_utils.plot_coastlines(basemap_object=basemap_object,
axes_object=axes_object,
line_colour=BORDER_COLOUR)
plotting_utils.plot_countries(basemap_object=basemap_object,
axes_object=axes_object,
line_colour=BORDER_COLOUR)
plotting_utils.plot_states_and_provinces(basemap_object=basemap_object,
axes_object=axes_object,
line_colour=BORDER_COLOUR)
plotting_utils.plot_parallels(basemap_object=basemap_object,
axes_object=axes_object,
num_parallels=NUM_PARALLELS)
plotting_utils.plot_meridians(basemap_object=basemap_object,
axes_object=axes_object,
num_meridians=NUM_MERIDIANS)
radar_plotting.plot_latlng_grid(
field_matrix=reflectivity_matrix_dbz,
field_name=RADAR_FIELD_NAME,
axes_object=axes_object,
min_grid_point_latitude_deg=numpy.min(latitudes_deg),
min_grid_point_longitude_deg=numpy.min(longitudes_deg),
latitude_spacing_deg=latitudes_deg[1] - latitudes_deg[0],
longitude_spacing_deg=longitudes_deg[1] - longitudes_deg[0])
if add_colour_bar:
colour_map_object, colour_norm_object = (
radar_plotting.get_default_colour_scheme(RADAR_FIELD_NAME))
plotting_utils.plot_colour_bar(axes_object_or_matrix=axes_object,
data_matrix=reflectivity_matrix_dbz,
colour_map_object=colour_map_object,
colour_norm_object=colour_norm_object,
orientation_string='horizontal',
padding=0.05,
extend_min=False,
extend_max=True,
fraction_of_axis_length=1.)
plotting_utils.label_axes(axes_object=axes_object,
label_string='({0:s})'.format(panel_letter),
y_coord_normalized=1.03)
print('Saving figure to: "{0:s}"...'.format(output_file_name))
figure_object.savefig(output_file_name,
dpi=FIGURE_RESOLUTION_DPI,
pad_inches=0,
bbox_inches='tight')
pyplot.close(figure_object)
def _plot_one_example_one_time(storm_object_table, full_id_string,
valid_time_unix_sec, tornado_table,
top_myrorss_dir_name, radar_field_name,
radar_height_m_asl, latitude_limits_deg,
longitude_limits_deg):
"""Plots one example with surrounding context at one time.
:param storm_object_table: pandas DataFrame, containing only storm objects
at one time with the relevant primary ID. Columns are documented in
`storm_tracking_io.write_file`.
:param full_id_string: Full ID of storm of interest.
:param valid_time_unix_sec: Valid time.
:param tornado_table: pandas DataFrame created by
`linkage._read_input_tornado_reports`.
:param top_myrorss_dir_name: See documentation at top of file.
:param radar_field_name: Same.
:param radar_height_m_asl: Same.
:param latitude_limits_deg: See doc for `_get_plotting_limits`.
:param longitude_limits_deg: Same.
"""
min_plot_latitude_deg = latitude_limits_deg[0]
max_plot_latitude_deg = latitude_limits_deg[1]
min_plot_longitude_deg = longitude_limits_deg[0]
max_plot_longitude_deg = longitude_limits_deg[1]
radar_file_name = myrorss_and_mrms_io.find_raw_file(
top_directory_name=top_myrorss_dir_name,
spc_date_string=time_conversion.time_to_spc_date_string(
valid_time_unix_sec),
unix_time_sec=valid_time_unix_sec,
data_source=radar_utils.MYRORSS_SOURCE_ID,
field_name=radar_field_name,
height_m_asl=radar_height_m_asl,
raise_error_if_missing=True)
print('Reading data from: "{0:s}"...'.format(radar_file_name))
radar_metadata_dict = myrorss_and_mrms_io.read_metadata_from_raw_file(
netcdf_file_name=radar_file_name,
data_source=radar_utils.MYRORSS_SOURCE_ID)
sparse_grid_table = (myrorss_and_mrms_io.read_data_from_sparse_grid_file(
netcdf_file_name=radar_file_name,
field_name_orig=radar_metadata_dict[
myrorss_and_mrms_io.FIELD_NAME_COLUMN_ORIG],
data_source=radar_utils.MYRORSS_SOURCE_ID,
sentinel_values=radar_metadata_dict[radar_utils.SENTINEL_VALUE_COLUMN])
)
radar_matrix, grid_point_latitudes_deg, grid_point_longitudes_deg = (
radar_s2f.sparse_to_full_grid(sparse_grid_table=sparse_grid_table,
metadata_dict=radar_metadata_dict))
radar_matrix = numpy.flip(radar_matrix, axis=0)
grid_point_latitudes_deg = grid_point_latitudes_deg[::-1]
axes_object, basemap_object = (
plotting_utils.create_equidist_cylindrical_map(
min_latitude_deg=min_plot_latitude_deg,
max_latitude_deg=max_plot_latitude_deg,
min_longitude_deg=min_plot_longitude_deg,
max_longitude_deg=max_plot_longitude_deg,
resolution_string='i')[1:])
plotting_utils.plot_coastlines(basemap_object=basemap_object,
axes_object=axes_object,
line_colour=BORDER_COLOUR)
plotting_utils.plot_countries(basemap_object=basemap_object,
axes_object=axes_object,
line_colour=BORDER_COLOUR)
plotting_utils.plot_states_and_provinces(basemap_object=basemap_object,
axes_object=axes_object,
line_colour=BORDER_COLOUR)
plotting_utils.plot_parallels(basemap_object=basemap_object,
axes_object=axes_object,
num_parallels=NUM_PARALLELS)
plotting_utils.plot_meridians(basemap_object=basemap_object,
axes_object=axes_object,
num_meridians=NUM_MERIDIANS)
radar_plotting.plot_latlng_grid(
field_matrix=radar_matrix,
field_name=radar_field_name,
axes_object=axes_object,
min_grid_point_latitude_deg=numpy.min(grid_point_latitudes_deg),
min_grid_point_longitude_deg=numpy.min(grid_point_longitudes_deg),
latitude_spacing_deg=numpy.diff(grid_point_latitudes_deg[:2])[0],
longitude_spacing_deg=numpy.diff(grid_point_longitudes_deg[:2])[0])
colour_map_object, colour_norm_object = (
radar_plotting.get_default_colour_scheme(radar_field_name))
plotting_utils.plot_colour_bar(axes_object_or_matrix=axes_object,
data_matrix=radar_matrix,
colour_map_object=colour_map_object,
colour_norm_object=colour_norm_object,
orientation_string='horizontal',
extend_min=False,
extend_max=True,
fraction_of_axis_length=0.8)
first_list, second_list = temporal_tracking.full_to_partial_ids(
[full_id_string])
primary_id_string = first_list[0]
secondary_id_string = second_list[0]
# Plot outlines of unrelated storms (with different primary IDs).
this_storm_object_table = storm_object_table.loc[storm_object_table[
tracking_utils.PRIMARY_ID_COLUMN] != primary_id_string]
storm_plotting.plot_storm_outlines(
storm_object_table=this_storm_object_table,
axes_object=axes_object,
basemap_object=basemap_object,
line_width=2,
line_colour='k',
line_style='dashed')
# Plot outlines of related storms (with the same primary ID).
this_storm_object_table = storm_object_table.loc[
(storm_object_table[tracking_utils.PRIMARY_ID_COLUMN] ==
primary_id_string) & (storm_object_table[
tracking_utils.SECONDARY_ID_COLUMN] != secondary_id_string)]
this_num_storm_objects = len(this_storm_object_table.index)
if this_num_storm_objects > 0:
storm_plotting.plot_storm_outlines(
storm_object_table=this_storm_object_table,
axes_object=axes_object,
basemap_object=basemap_object,
line_width=2,
line_colour='k',
line_style='solid')
for j in range(len(this_storm_object_table)):
axes_object.text(
this_storm_object_table[
tracking_utils.CENTROID_LONGITUDE_COLUMN].values[j],
this_storm_object_table[
tracking_utils.CENTROID_LATITUDE_COLUMN].values[j],
'P',
fontsize=FONT_SIZE,
color=FONT_COLOUR,
fontweight='bold',
horizontalalignment='center',
verticalalignment='center')
# Plot outline of storm of interest (same secondary ID).
this_storm_object_table = storm_object_table.loc[storm_object_table[
tracking_utils.SECONDARY_ID_COLUMN] == secondary_id_string]
storm_plotting.plot_storm_outlines(
storm_object_table=this_storm_object_table,
axes_object=axes_object,
basemap_object=basemap_object,
line_width=4,
line_colour='k',
line_style='solid')
this_num_storm_objects = len(this_storm_object_table.index)
plot_forecast = (this_num_storm_objects > 0 and FORECAST_PROBABILITY_COLUMN
in list(this_storm_object_table))
if plot_forecast:
this_polygon_object_latlng = this_storm_object_table[
tracking_utils.LATLNG_POLYGON_COLUMN].values[0]
this_latitude_deg = numpy.min(
numpy.array(this_polygon_object_latlng.exterior.xy[1]))
this_longitude_deg = this_storm_object_table[
tracking_utils.CENTROID_LONGITUDE_COLUMN].values[0]
label_string = 'Prob = {0:.3f}\nat {1:s}'.format(
this_storm_object_table[FORECAST_PROBABILITY_COLUMN].values[0],
time_conversion.unix_sec_to_string(valid_time_unix_sec,
TORNADO_TIME_FORMAT))
bounding_box_dict = {
'facecolor':
plotting_utils.colour_from_numpy_to_tuple(
PROBABILITY_BACKGROUND_COLOUR),
'alpha':
PROBABILITY_BACKGROUND_OPACITY,
'edgecolor':
'k',
'linewidth':
1
}
axes_object.text(this_longitude_deg,
this_latitude_deg,
label_string,
fontsize=FONT_SIZE,
color=plotting_utils.colour_from_numpy_to_tuple(
PROBABILITY_FONT_COLOUR),
fontweight='bold',
bbox=bounding_box_dict,
horizontalalignment='center',
verticalalignment='top',
zorder=1e10)
tornado_latitudes_deg = tornado_table[linkage.EVENT_LATITUDE_COLUMN].values
tornado_longitudes_deg = tornado_table[
linkage.EVENT_LONGITUDE_COLUMN].values
tornado_times_unix_sec = tornado_table[linkage.EVENT_TIME_COLUMN].values
tornado_time_strings = [
time_conversion.unix_sec_to_string(t, TORNADO_TIME_FORMAT)
for t in tornado_times_unix_sec
]
axes_object.plot(tornado_longitudes_deg,
tornado_latitudes_deg,
linestyle='None',
marker=TORNADO_MARKER_TYPE,
markersize=TORNADO_MARKER_SIZE,
markeredgewidth=TORNADO_MARKER_EDGE_WIDTH,
markerfacecolor=plotting_utils.colour_from_numpy_to_tuple(
TORNADO_MARKER_COLOUR),
markeredgecolor=plotting_utils.colour_from_numpy_to_tuple(
TORNADO_MARKER_COLOUR))
num_tornadoes = len(tornado_latitudes_deg)
for j in range(num_tornadoes):
axes_object.text(tornado_longitudes_deg[j] + 0.02,
tornado_latitudes_deg[j] - 0.02,
tornado_time_strings[j],
fontsize=FONT_SIZE,
color=FONT_COLOUR,
fontweight='bold',
horizontalalignment='left',
verticalalignment='top')
def _plot_2d3d_radar_scan(list_of_predictor_matrices,
model_metadata_dict,
allow_whitespace,
title_string=None):
"""Plots 3-D reflectivity and 2-D azimuthal shear for one example.
:param list_of_predictor_matrices: See doc for `_plot_3d_radar_scan`.
:param model_metadata_dict: Same.
:param allow_whitespace: Same.
:param title_string: Same.
:return: figure_objects: length-2 list of figure handles (instances of
`matplotlib.figure.Figure`). The first is for reflectivity; the second
is for azimuthal shear.
:return: axes_object_matrices: length-2 list (the first is for reflectivity;
the second is for azimuthal shear). Each element is a 2-D numpy
array of axes handles (instances of
`matplotlib.axes._subplots.AxesSubplot`).
"""
training_option_dict = model_metadata_dict[cnn.TRAINING_OPTION_DICT_KEY]
az_shear_field_names = training_option_dict[trainval_io.RADAR_FIELDS_KEY]
refl_heights_m_agl = training_option_dict[trainval_io.RADAR_HEIGHTS_KEY]
num_az_shear_fields = len(az_shear_field_names)
num_refl_heights = len(refl_heights_m_agl)
this_num_panel_rows = int(numpy.floor(numpy.sqrt(num_refl_heights)))
this_num_panel_columns = int(
numpy.ceil(float(num_refl_heights) / this_num_panel_rows))
if allow_whitespace:
refl_figure_object = None
refl_axes_object_matrix = None
else:
refl_figure_object, refl_axes_object_matrix = (
plotting_utils.create_paneled_figure(
num_rows=this_num_panel_rows,
num_columns=this_num_panel_columns,
horizontal_spacing=0.,
vertical_spacing=0.,
shared_x_axis=False,
shared_y_axis=False,
keep_aspect_ratio=True))
refl_figure_object, refl_axes_object_matrix = (
radar_plotting.plot_3d_grid_without_coords(
field_matrix=numpy.flip(list_of_predictor_matrices[0][..., 0],
axis=0),
field_name=radar_utils.REFL_NAME,
grid_point_heights_metres=refl_heights_m_agl,
ground_relative=True,
num_panel_rows=this_num_panel_rows,
figure_object=refl_figure_object,
axes_object_matrix=refl_axes_object_matrix,
font_size=FONT_SIZE_SANS_COLOUR_BARS))
if allow_whitespace:
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(radar_utils.REFL_NAME))
plotting_utils.plot_colour_bar(
axes_object_or_matrix=refl_axes_object_matrix,
data_matrix=list_of_predictor_matrices[0],
colour_map_object=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation_string='horizontal',
extend_min=True,
extend_max=True)
if title_string is not None:
this_title_string = '{0:s}; {1:s}'.format(title_string,
radar_utils.REFL_NAME)
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
if allow_whitespace:
shear_figure_object = None
shear_axes_object_matrix = None
else:
shear_figure_object, shear_axes_object_matrix = (
plotting_utils.create_paneled_figure(
num_rows=1,
num_columns=num_az_shear_fields,
horizontal_spacing=0.,
vertical_spacing=0.,
shared_x_axis=False,
shared_y_axis=False,
keep_aspect_ratio=True))
shear_figure_object, shear_axes_object_matrix = (
radar_plotting.plot_many_2d_grids_without_coords(
field_matrix=numpy.flip(list_of_predictor_matrices[1], axis=0),
field_name_by_panel=az_shear_field_names,
panel_names=az_shear_field_names,
num_panel_rows=1,
figure_object=shear_figure_object,
axes_object_matrix=shear_axes_object_matrix,
plot_colour_bar_by_panel=numpy.full(num_az_shear_fields,
False,
dtype=bool),
font_size=FONT_SIZE_SANS_COLOUR_BARS))
if allow_whitespace:
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(
radar_utils.LOW_LEVEL_SHEAR_NAME))
plotting_utils.plot_colour_bar(
axes_object_or_matrix=shear_axes_object_matrix,
data_matrix=list_of_predictor_matrices[1],
colour_map_object=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation_string='horizontal',
extend_min=True,
extend_max=True)
if title_string is not None:
pyplot.suptitle(title_string, fontsize=TITLE_FONT_SIZE)
figure_objects = [refl_figure_object, shear_figure_object]
axes_object_matrices = [refl_axes_object_matrix, shear_axes_object_matrix]
return figure_objects, axes_object_matrices
def _plot_3d_radar_scan(list_of_predictor_matrices,
model_metadata_dict,
allow_whitespace,
title_string=None):
"""Plots 3-D radar scan for one example.
J = number of panel rows in image
K = number of panel columns in image
F = number of radar fields
:param list_of_predictor_matrices: List created by
`testing_io.read_specific_examples`, except that the first axis (example
dimension) is removed.
:param model_metadata_dict: Dictionary returned by
`cnn.read_model_metadata`.
:param allow_whitespace: See documentation at top of file.
:param title_string: Title (may be None).
:return: figure_objects: length-F list of figure handles (instances of
`matplotlib.figure.Figure`).
:return: axes_object_matrices: length-F list. Each element is a J-by-K
numpy array of axes handles (instances of
`matplotlib.axes._subplots.AxesSubplot`).
"""
training_option_dict = model_metadata_dict[cnn.TRAINING_OPTION_DICT_KEY]
radar_field_names = training_option_dict[trainval_io.RADAR_FIELDS_KEY]
radar_heights_m_agl = training_option_dict[trainval_io.RADAR_HEIGHTS_KEY]
num_radar_fields = len(radar_field_names)
num_radar_heights = len(radar_heights_m_agl)
num_panel_rows = int(numpy.floor(numpy.sqrt(num_radar_heights)))
num_panel_columns = int(
numpy.ceil(float(num_radar_heights) / num_panel_rows))
figure_objects = [None] * num_radar_fields
axes_object_matrices = [None] * num_radar_fields
radar_matrix = list_of_predictor_matrices[0]
for j in range(num_radar_fields):
this_radar_matrix = numpy.flip(radar_matrix[..., j], axis=0)
if not allow_whitespace:
figure_objects[j], axes_object_matrices[j] = (
plotting_utils.create_paneled_figure(
num_rows=num_panel_rows,
num_columns=num_panel_columns,
horizontal_spacing=0.,
vertical_spacing=0.,
shared_x_axis=False,
shared_y_axis=False,
keep_aspect_ratio=True))
figure_objects[j], axes_object_matrices[j] = (
radar_plotting.plot_3d_grid_without_coords(
field_matrix=this_radar_matrix,
field_name=radar_field_names[j],
grid_point_heights_metres=radar_heights_m_agl,
ground_relative=True,
num_panel_rows=num_panel_rows,
figure_object=figure_objects[j],
axes_object_matrix=axes_object_matrices[j],
font_size=FONT_SIZE_SANS_COLOUR_BARS))
if allow_whitespace:
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(radar_field_names[j]))
plotting_utils.plot_colour_bar(
axes_object_or_matrix=axes_object_matrices[j],
data_matrix=this_radar_matrix,
colour_map_object=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation_string='horizontal',
extend_min=True,
extend_max=True)
if title_string is not None:
this_title_string = '{0:s}; {1:s}'.format(
title_string, radar_field_names[j])
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
return figure_objects, axes_object_matrices
def _plot_one_example_one_time(
storm_object_table, full_id_string, valid_time_unix_sec,
tornado_table, top_myrorss_dir_name, radar_field_name,
radar_height_m_asl, latitude_limits_deg, longitude_limits_deg):
"""Plots one example with surrounding context at one time.
:param storm_object_table: pandas DataFrame, containing only storm objects
at one time with the relevant primary ID. Columns are documented in
`storm_tracking_io.write_file`.
:param full_id_string: Full ID of storm of interest.
:param valid_time_unix_sec: Valid time.
:param tornado_table: pandas DataFrame created by
`linkage._read_input_tornado_reports`.
:param top_myrorss_dir_name: See documentation at top of file.
:param radar_field_name: Same.
:param radar_height_m_asl: Same.
:param latitude_limits_deg: See doc for `_get_plotting_limits`.
:param longitude_limits_deg: Same.
"""
min_plot_latitude_deg = latitude_limits_deg[0]
max_plot_latitude_deg = latitude_limits_deg[1]
min_plot_longitude_deg = longitude_limits_deg[0]
max_plot_longitude_deg = longitude_limits_deg[1]
radar_file_name = myrorss_and_mrms_io.find_raw_file_inexact_time(
top_directory_name=top_myrorss_dir_name,
desired_time_unix_sec=valid_time_unix_sec,
spc_date_string=time_conversion.time_to_spc_date_string(
valid_time_unix_sec),
data_source=radar_utils.MYRORSS_SOURCE_ID,
field_name=radar_field_name, height_m_asl=radar_height_m_asl,
max_time_offset_sec=
myrorss_and_mrms_io.DEFAULT_MAX_TIME_OFFSET_FOR_NON_SHEAR_SEC,
raise_error_if_missing=True)
print('Reading data from: "{0:s}"...'.format(radar_file_name))
radar_metadata_dict = myrorss_and_mrms_io.read_metadata_from_raw_file(
netcdf_file_name=radar_file_name,
data_source=radar_utils.MYRORSS_SOURCE_ID)
sparse_grid_table = (
myrorss_and_mrms_io.read_data_from_sparse_grid_file(
netcdf_file_name=radar_file_name,
field_name_orig=radar_metadata_dict[
myrorss_and_mrms_io.FIELD_NAME_COLUMN_ORIG],
data_source=radar_utils.MYRORSS_SOURCE_ID,
sentinel_values=radar_metadata_dict[
radar_utils.SENTINEL_VALUE_COLUMN]
)
)
radar_matrix, grid_point_latitudes_deg, grid_point_longitudes_deg = (
radar_s2f.sparse_to_full_grid(
sparse_grid_table=sparse_grid_table,
metadata_dict=radar_metadata_dict)
)
radar_matrix = numpy.flip(radar_matrix, axis=0)
grid_point_latitudes_deg = grid_point_latitudes_deg[::-1]
axes_object, basemap_object = (
plotting_utils.create_equidist_cylindrical_map(
min_latitude_deg=min_plot_latitude_deg,
max_latitude_deg=max_plot_latitude_deg,
min_longitude_deg=min_plot_longitude_deg,
max_longitude_deg=max_plot_longitude_deg, resolution_string='h'
)[1:]
)
plotting_utils.plot_coastlines(
basemap_object=basemap_object, axes_object=axes_object,
line_colour=plotting_utils.DEFAULT_COUNTRY_COLOUR)
plotting_utils.plot_countries(
basemap_object=basemap_object, axes_object=axes_object)
plotting_utils.plot_states_and_provinces(
basemap_object=basemap_object, axes_object=axes_object)
plotting_utils.plot_parallels(
basemap_object=basemap_object, axes_object=axes_object,
num_parallels=NUM_PARALLELS, line_width=0)
plotting_utils.plot_meridians(
basemap_object=basemap_object, axes_object=axes_object,
num_meridians=NUM_MERIDIANS, line_width=0)
radar_plotting.plot_latlng_grid(
field_matrix=radar_matrix, field_name=radar_field_name,
axes_object=axes_object,
min_grid_point_latitude_deg=numpy.min(grid_point_latitudes_deg),
min_grid_point_longitude_deg=numpy.min(grid_point_longitudes_deg),
latitude_spacing_deg=numpy.diff(grid_point_latitudes_deg[:2])[0],
longitude_spacing_deg=numpy.diff(grid_point_longitudes_deg[:2])[0]
)
colour_map_object, colour_norm_object = (
radar_plotting.get_default_colour_scheme(radar_field_name)
)
plotting_utils.plot_colour_bar(
axes_object_or_matrix=axes_object, data_matrix=radar_matrix,
colour_map_object=colour_map_object,
colour_norm_object=colour_norm_object, orientation_string='horizontal',
padding=0.05, extend_min=False, extend_max=True,
fraction_of_axis_length=0.8)
first_list, second_list = temporal_tracking.full_to_partial_ids(
[full_id_string]
)
primary_id_string = first_list[0]
secondary_id_string = second_list[0]
# Plot outlines of unrelated storms (with different primary IDs).
this_storm_object_table = storm_object_table.loc[
storm_object_table[tracking_utils.PRIMARY_ID_COLUMN] !=
primary_id_string
]
storm_plotting.plot_storm_outlines(
storm_object_table=this_storm_object_table, axes_object=axes_object,
basemap_object=basemap_object, line_width=AUXILIARY_STORM_WIDTH,
line_colour='k', line_style='dashed')
# Plot outlines of related storms (with the same primary ID).
this_storm_object_table = storm_object_table.loc[
(storm_object_table[tracking_utils.PRIMARY_ID_COLUMN] ==
primary_id_string) &
(storm_object_table[tracking_utils.SECONDARY_ID_COLUMN] !=
secondary_id_string)
]
this_num_storm_objects = len(this_storm_object_table.index)
if this_num_storm_objects > 0:
storm_plotting.plot_storm_outlines(
storm_object_table=this_storm_object_table, axes_object=axes_object,
basemap_object=basemap_object, line_width=AUXILIARY_STORM_WIDTH,
line_colour='k', line_style='solid'
)
for j in range(len(this_storm_object_table)):
axes_object.text(
this_storm_object_table[
tracking_utils.CENTROID_LONGITUDE_COLUMN
].values[j],
this_storm_object_table[
tracking_utils.CENTROID_LATITUDE_COLUMN
].values[j],
'P',
fontsize=MAIN_FONT_SIZE, color=FONT_COLOUR, fontweight='bold',
horizontalalignment='center', verticalalignment='center'
)
# Plot outline of storm of interest (same secondary ID).
this_storm_object_table = storm_object_table.loc[
storm_object_table[tracking_utils.SECONDARY_ID_COLUMN] ==
secondary_id_string
]
storm_plotting.plot_storm_outlines(
storm_object_table=this_storm_object_table, axes_object=axes_object,
basemap_object=basemap_object, line_width=MAIN_STORM_WIDTH,
line_colour='k', line_style='solid')
this_num_storm_objects = len(this_storm_object_table.index)
plot_forecast = (
this_num_storm_objects > 0 and
FORECAST_PROBABILITY_COLUMN in list(this_storm_object_table)
)
if plot_forecast:
label_string = 'Prob = {0:.3f}\nat {1:s}'.format(
this_storm_object_table[FORECAST_PROBABILITY_COLUMN].values[0],
time_conversion.unix_sec_to_string(
valid_time_unix_sec, TORNADO_TIME_FORMAT)
)
axes_object.set_title(
label_string.replace('\n', ' '), fontsize=TITLE_FONT_SIZE
)
tornado_id_strings = tornado_table[tornado_io.TORNADO_ID_COLUMN].values
for this_tornado_id_string in numpy.unique(tornado_id_strings):
these_rows = numpy.where(
tornado_id_strings == this_tornado_id_string
)[0]
this_tornado_table = tornado_table.iloc[these_rows].sort_values(
linkage.EVENT_TIME_COLUMN, axis=0, ascending=True, inplace=False
)
_plot_one_tornado(
tornado_table=this_tornado_table, axes_object=axes_object
)
def plot_examples(list_of_predictor_matrices,
storm_ids,
storm_times_unix_sec,
model_metadata_dict,
output_dir_name,
storm_activations=None):
"""Plots one or more learning examples.
E = number of examples (storm objects)
:param list_of_predictor_matrices: List created by
`testing_io.read_specific_examples`. Contains data to be plotted.
:param storm_ids: length-E list of storm IDs.
:param storm_times_unix_sec: length-E numpy array of storm times.
:param model_metadata_dict: See doc for `cnn.read_model_metadata`.
:param output_dir_name: Name of output directory (figures will be saved
here).
:param storm_activations: length-E numpy array of storm activations (may be
None). Will be included in title of each figure.
"""
training_option_dict = model_metadata_dict[cnn.TRAINING_OPTION_DICT_KEY]
sounding_field_names = training_option_dict[
trainval_io.SOUNDING_FIELDS_KEY]
plot_soundings = sounding_field_names is not None
if plot_soundings:
list_of_metpy_dictionaries = dl_utils.soundings_to_metpy_dictionaries(
sounding_matrix=list_of_predictor_matrices[-1],
field_names=sounding_field_names)
else:
list_of_metpy_dictionaries = None
num_radar_dimensions = len(list_of_predictor_matrices[0].shape) - 2
list_of_layer_operation_dicts = model_metadata_dict[
cnn.LAYER_OPERATIONS_KEY]
if num_radar_dimensions == 2:
if list_of_layer_operation_dicts is None:
field_name_by_panel = training_option_dict[
trainval_io.RADAR_FIELDS_KEY]
panel_names = (
radar_plotting.radar_fields_and_heights_to_panel_names(
field_names=field_name_by_panel,
heights_m_agl=training_option_dict[
trainval_io.RADAR_HEIGHTS_KEY]))
plot_colour_bar_by_panel = numpy.full(len(panel_names),
True,
dtype=bool)
else:
field_name_by_panel, panel_names = (
radar_plotting.layer_ops_to_field_and_panel_names(
list_of_layer_operation_dicts))
plot_colour_bar_by_panel = numpy.full(len(panel_names),
False,
dtype=bool)
plot_colour_bar_by_panel[2::3] = True
else:
field_name_by_panel = None
panel_names = None
plot_colour_bar_by_panel = None
az_shear_field_names = training_option_dict[trainval_io.RADAR_FIELDS_KEY]
num_az_shear_fields = len(az_shear_field_names)
num_storms = len(storm_ids)
myrorss_2d3d = len(list_of_predictor_matrices) == 3
for i in range(num_storms):
this_time_string = time_conversion.unix_sec_to_string(
storm_times_unix_sec[i], TIME_FORMAT)
this_base_title_string = 'Storm "{0:s}" at {1:s}'.format(
storm_ids[i], this_time_string)
if storm_activations is not None:
this_base_title_string += ' (activation = {0:.3f})'.format(
storm_activations[i])
this_base_file_name = '{0:s}/storm={1:s}_{2:s}'.format(
output_dir_name, storm_ids[i].replace('_', '-'), this_time_string)
if plot_soundings:
sounding_plotting.plot_sounding(
sounding_dict_for_metpy=list_of_metpy_dictionaries[i],
title_string=this_base_title_string)
this_file_name = '{0:s}_sounding.jpg'.format(this_base_file_name)
print 'Saving figure to: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
if myrorss_2d3d:
this_reflectivity_matrix_dbz = numpy.flip(
list_of_predictor_matrices[0][i, ..., 0], axis=0)
this_num_heights = this_reflectivity_matrix_dbz.shape[-1]
this_num_panel_rows = int(numpy.floor(
numpy.sqrt(this_num_heights)))
_, these_axes_objects = radar_plotting.plot_3d_grid_without_coords(
field_matrix=this_reflectivity_matrix_dbz,
field_name=radar_utils.REFL_NAME,
grid_point_heights_metres=training_option_dict[
trainval_io.RADAR_HEIGHTS_KEY],
ground_relative=True,
num_panel_rows=this_num_panel_rows,
font_size=FONT_SIZE_SANS_COLOUR_BARS)
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(
radar_utils.REFL_NAME))
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=this_reflectivity_matrix_dbz,
colour_map=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal',
extend_min=True,
extend_max=True)
this_title_string = '{0:s}; {1:s}'.format(this_base_title_string,
radar_utils.REFL_NAME)
this_file_name = '{0:s}_reflectivity.jpg'.format(
this_base_file_name)
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
this_az_shear_matrix_s01 = numpy.flip(
list_of_predictor_matrices[1][i, ..., 0], axis=0)
_, these_axes_objects = (
radar_plotting.plot_many_2d_grids_without_coords(
field_matrix=this_az_shear_matrix_s01,
field_name_by_panel=az_shear_field_names,
panel_names=az_shear_field_names,
num_panel_rows=1,
plot_colour_bar_by_panel=numpy.full(num_az_shear_fields,
False,
dtype=bool),
font_size=FONT_SIZE_SANS_COLOUR_BARS))
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(
radar_utils.LOW_LEVEL_SHEAR_NAME))
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=this_az_shear_matrix_s01,
colour_map=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal',
extend_min=True,
extend_max=True)
this_file_name = '{0:s}_shear.jpg'.format(this_base_file_name)
pyplot.suptitle(this_base_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
continue
this_radar_matrix = list_of_predictor_matrices[0]
if num_radar_dimensions == 2:
this_num_channels = this_radar_matrix.shape[-1]
this_num_panel_rows = int(
numpy.floor(numpy.sqrt(this_num_channels)))
radar_plotting.plot_many_2d_grids_without_coords(
field_matrix=numpy.flip(this_radar_matrix[i, ...], axis=0),
field_name_by_panel=field_name_by_panel,
panel_names=panel_names,
num_panel_rows=this_num_panel_rows,
plot_colour_bar_by_panel=plot_colour_bar_by_panel,
font_size=FONT_SIZE_WITH_COLOUR_BARS,
row_major=False)
this_title_string = this_base_title_string + ''
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
this_file_name = '{0:s}.jpg'.format(this_base_file_name)
print 'Saving figure to: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
continue
radar_field_names = training_option_dict[trainval_io.RADAR_FIELDS_KEY]
radar_heights_m_agl = training_option_dict[
trainval_io.RADAR_HEIGHTS_KEY]
for j in range(len(radar_field_names)):
this_num_heights = this_radar_matrix.shape[-2]
this_num_panel_rows = int(numpy.floor(
numpy.sqrt(this_num_heights)))
_, these_axes_objects = radar_plotting.plot_3d_grid_without_coords(
field_matrix=numpy.flip(this_radar_matrix[i, ..., j], axis=0),
field_name=radar_field_names[j],
grid_point_heights_metres=radar_heights_m_agl,
ground_relative=True,
num_panel_rows=this_num_panel_rows,
font_size=FONT_SIZE_SANS_COLOUR_BARS)
this_colour_map_object, this_colour_norm_object = (
radar_plotting.get_default_colour_scheme(radar_field_names[j]))
plotting_utils.add_colour_bar(
axes_object_or_list=these_axes_objects,
values_to_colour=this_radar_matrix[i, ..., j],
colour_map=this_colour_map_object,
colour_norm_object=this_colour_norm_object,
orientation='horizontal',
extend_min=True,
extend_max=True)
this_title_string = '{0:s}; {1:s}'.format(this_base_title_string,
radar_field_names[j])
this_file_name = '{0:s}_{1:s}.jpg'.format(
this_base_file_name, radar_field_names[j].replace('_', '-'))
pyplot.suptitle(this_title_string, fontsize=TITLE_FONT_SIZE)
print 'Saving figure to: "{0:s}"...'.format(this_file_name)
pyplot.savefig(this_file_name, dpi=FIGURE_RESOLUTION_DPI)
pyplot.close()
def _plot_echo_tops(echo_top_matrix_km_asl, latitudes_deg, longitudes_deg,
plot_colour_bar, convective_flag_matrix=None):
"""Plots grid of 40-dBZ echo tops.
M = number of rows in grid
N = number of columns in grid
:param echo_top_matrix_km_asl: M-by-N numpy array of echo tops (km above sea
level).
:param latitudes_deg: length-M numpy array of latitudes (deg N).
:param longitudes_deg: length-N numpy array of longitudes (deg E).
:param plot_colour_bar: Boolean flag.
:param convective_flag_matrix: M-by-N numpy array of Boolean flags,
indicating which grid cells are convective. If
`convective_flag_matrix is None`, all grid cells will be plotted. If
`convective_flag_matrix is not None`, only convective grid cells will be
plotted.
:return: figure_object: Figure handle (instance of
`matplotlib.figure.Figure`).
:return: axes_object: Axes handle (instance of
`matplotlib.axes._subplots.AxesSubplot`).
:return: basemap_object: Basemap handle (instance of
`mpl_toolkits.basemap.Basemap`).
"""
figure_object, axes_object, basemap_object = (
plotting_utils.create_equidist_cylindrical_map(
min_latitude_deg=numpy.min(latitudes_deg),
max_latitude_deg=numpy.max(latitudes_deg),
min_longitude_deg=numpy.min(longitudes_deg),
max_longitude_deg=numpy.max(longitudes_deg), resolution_string='h'
)
)
# plotting_utils.plot_coastlines(
# basemap_object=basemap_object, axes_object=axes_object,
# line_colour=plotting_utils.DEFAULT_COUNTRY_COLOUR
# )
plotting_utils.plot_countries(
basemap_object=basemap_object, axes_object=axes_object
)
plotting_utils.plot_states_and_provinces(
basemap_object=basemap_object, axes_object=axes_object
)
plotting_utils.plot_parallels(
basemap_object=basemap_object, axes_object=axes_object,
num_parallels=NUM_PARALLELS, line_width=0
)
plotting_utils.plot_meridians(
basemap_object=basemap_object, axes_object=axes_object,
num_meridians=NUM_MERIDIANS, line_width=0
)
matrix_to_plot = echo_top_matrix_km_asl + 0.
if convective_flag_matrix is not None:
matrix_to_plot[convective_flag_matrix == False] = numpy.nan
radar_plotting.plot_latlng_grid(
field_matrix=matrix_to_plot, field_name=radar_utils.ECHO_TOP_40DBZ_NAME,
axes_object=axes_object,
min_grid_point_latitude_deg=numpy.min(latitudes_deg),
min_grid_point_longitude_deg=numpy.min(longitudes_deg),
latitude_spacing_deg=numpy.diff(latitudes_deg[:2])[0],
longitude_spacing_deg=numpy.diff(longitudes_deg[:2])[0]
)
if not plot_colour_bar:
return figure_object, axes_object, basemap_object
colour_map_object, colour_norm_object = (
radar_plotting.get_default_colour_scheme(
radar_utils.ECHO_TOP_40DBZ_NAME)
)
colour_bar_object = plotting_utils.plot_colour_bar(
axes_object_or_matrix=axes_object, data_matrix=matrix_to_plot,
colour_map_object=colour_map_object,
colour_norm_object=colour_norm_object, orientation_string='horizontal',
extend_min=False, extend_max=True, fraction_of_axis_length=1.
)
colour_bar_object.set_label('40-dBZ echo top (kft ASL)')
return figure_object, axes_object, basemap_object
