def test_find_predecessors_first_5sec_all_0changes(self):
"""Ensures correct output from find_predecessors.
In this case, working on the first table with max time difference =
5 seconds and max ID changes = 0. Looking for all predecessors on path
to earliest.
"""
this_num_storm_objects = len(FIRST_STORM_OBJECT_TABLE.index)
for i in range(this_num_storm_objects):
these_predecessor_rows = temporal_tracking.find_predecessors(
storm_object_table=FIRST_STORM_OBJECT_TABLE,
target_row=i,
num_seconds_back=5,
max_num_sec_id_changes=0,
return_all_on_path=True)
these_predecessor_rows = numpy.sort(these_predecessor_rows)
these_expected_rows = numpy.sort(
numpy.array(FIRST_PREDECESSOR_DICT_5SEC_ALL_0CHANGES[i],
dtype=int))
self.assertTrue(
numpy.array_equal(these_predecessor_rows, these_expected_rows))
def test_find_predecessors_first_5sec(self):
"""Ensures correct output from find_predecessors.
In this case, working on the first table with max time difference =
5 seconds.
"""
this_num_storm_objects = len(FIRST_STORM_OBJECT_TABLE.index)
for i in range(this_num_storm_objects):
these_predecessor_rows = temporal_tracking.find_predecessors(
storm_object_table=FIRST_STORM_OBJECT_TABLE,
target_row=i,
num_seconds_back=5)
these_predecessor_rows = numpy.sort(these_predecessor_rows)
these_expected_rows = numpy.sort(
numpy.array(FIRST_PREDECESSOR_DICT_5SEC[i], dtype=int))
self.assertTrue(
numpy.array_equal(these_predecessor_rows, these_expected_rows))
def _plot_schema(storm_object_table, output_file_name):
"""Plots schema for storm-velocity estimation.
:param storm_object_table: pandas DataFrame created by
`_create_tracking_data`.
:param output_file_name: Path to output file (figure will be saved here).
"""
centroid_x_coords = storm_object_table[
tracking_utils.CENTROID_X_COLUMN].values
centroid_y_coords = storm_object_table[
tracking_utils.CENTROID_Y_COLUMN].values
secondary_id_strings = storm_object_table[
tracking_utils.SECONDARY_ID_COLUMN].values
storm_object_table = storm_object_table.assign(
**{
tracking_utils.CENTROID_LONGITUDE_COLUMN: centroid_x_coords,
tracking_utils.CENTROID_LATITUDE_COLUMN: centroid_y_coords
})
figure_object, axes_object, basemap_object = (
plotting_utils.create_equidist_cylindrical_map(
min_latitude_deg=numpy.min(centroid_y_coords),
max_latitude_deg=numpy.max(centroid_y_coords),
min_longitude_deg=numpy.min(centroid_x_coords),
max_longitude_deg=numpy.max(centroid_x_coords)))
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=TRACK_COLOUR,
line_width=TRACK_WIDTH,
start_marker_type=None,
end_marker_type=None)
num_storm_objects = len(storm_object_table.index)
predecessor_rows = temporal_tracking.find_predecessors(
storm_object_table=storm_object_table,
target_row=num_storm_objects - 1,
num_seconds_back=100,
return_all_on_path=False)
legend_handles = [None] * 3
legend_strings = [None] * 3
this_handle = axes_object.plot(centroid_x_coords[[0, 0]],
centroid_y_coords[[0, 0]],
color=TRACK_COLOUR,
linestyle='solid',
linewidth=TRACK_WIDTH)[0]
legend_handles[-1] = this_handle
legend_strings[-1] = 'Storm track'
for i in range(num_storm_objects):
if i in predecessor_rows or i == num_storm_objects - 1:
this_colour = SPECIAL_STORM_COLOUR
else:
this_colour = DEFAULT_STORM_COLOUR
this_handle = axes_object.plot(centroid_x_coords[i],
centroid_y_coords[i],
linestyle='None',
marker=MARKER_TYPE,
markersize=MARKER_SIZE,
markerfacecolor=this_colour,
markeredgecolor=this_colour,
markeredgewidth=MARKER_EDGE_WIDTH)[0]
if i in predecessor_rows or i == num_storm_objects - 1:
legend_handles[0] = this_handle
legend_strings[0] = 'Object used in\nvelocity estimate'
else:
legend_handles[1] = this_handle
legend_strings[1] = 'Object not used'
axes_object.text(centroid_x_coords[i],
centroid_y_coords[i] - TEXT_OFFSET,
secondary_id_strings[i],
color=this_colour,
fontsize=FONT_SIZE,
fontweight='bold',
horizontalalignment='center',
verticalalignment='top')
axes_object.set_yticks([], [])
storm_times_minutes = storm_object_table[
tracking_utils.VALID_TIME_COLUMN].values
x_tick_values, unique_indices = numpy.unique(centroid_x_coords,
return_index=True)
x_tick_labels = [
'{0:d}'.format(storm_times_minutes[i]) for i in unique_indices
]
axes_object.set_xticks(x_tick_values)
axes_object.set_xticklabels(x_tick_labels)
axes_object.set_xlabel('Time (minutes)')
axes_object.legend(legend_handles,
legend_strings,
fontsize=FONT_SIZE,
loc=(0.02, 0.55))
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 _handle_one_storm_cell(storm_object_table, primary_id_string,
conus_latitudes_deg, conus_longitudes_deg,
max_lead_time_sec):
"""Handles (either keeps or removes) one storm cell.
In this case, a "storm cell" is a group of storm objects with the same
primary ID.
V = number of vertices in CONUS boundary
:param storm_object_table: pandas DataFrame with columns listed in doc for
`storm_tracking_io.write_file`.
:param primary_id_string: Primary ID of storm cell.
:param conus_latitudes_deg: length-V numpy array of latitudes (deg N) in
boundary.
:param conus_longitudes_deg: length-V numpy array of longitudes (deg E) in
boundary.
:param max_lead_time_sec: See documentation at top of file.
:return: bad_object_indices: 1-D numpy array with indices of bad storm
objects (those with successor outside CONUS). These are row indices for
`storm_object_table`.
"""
object_in_cell_indices = numpy.where(storm_object_table[
tracking_utils.PRIMARY_ID_COLUMN].values == primary_id_string)[0]
query_latitudes_deg = []
query_longitudes_deg = []
query_object_indices = []
num_storm_objects = len(object_in_cell_indices)
for i in range(num_storm_objects):
j = object_in_cell_indices[i]
this_polygon_object = (
storm_object_table[tracking_utils.LATLNG_POLYGON_COLUMN].values[j])
these_latitudes_deg = numpy.array(this_polygon_object.exterior.xy[1])
these_longitudes_deg = numpy.array(this_polygon_object.exterior.xy[0])
# Find northeasternmost point in storm boundary.
this_index = numpy.argmax(these_latitudes_deg + these_longitudes_deg)
query_latitudes_deg.append(these_latitudes_deg[this_index])
query_longitudes_deg.append(these_longitudes_deg[this_index])
query_object_indices.append(j)
# Find southwesternmost point in storm boundary.
this_index = numpy.argmin(these_latitudes_deg + these_longitudes_deg)
query_latitudes_deg.append(these_latitudes_deg[this_index])
query_longitudes_deg.append(these_longitudes_deg[this_index])
query_object_indices.append(j)
# Find northwesternmost point in storm boundary.
this_index = numpy.argmax(these_latitudes_deg - these_longitudes_deg)
query_latitudes_deg.append(these_latitudes_deg[this_index])
query_longitudes_deg.append(these_longitudes_deg[this_index])
query_object_indices.append(j)
# Find northeasternmost point in storm boundary.
this_index = numpy.argmax(these_longitudes_deg - these_latitudes_deg)
query_latitudes_deg.append(these_latitudes_deg[this_index])
query_longitudes_deg.append(these_longitudes_deg[this_index])
query_object_indices.append(j)
query_latitudes_deg = numpy.array(query_latitudes_deg)
query_longitudes_deg = numpy.array(query_longitudes_deg)
query_object_indices = numpy.array(query_object_indices, dtype=int)
in_conus_flags = conus_boundary.find_points_in_conus(
conus_latitudes_deg=conus_latitudes_deg,
conus_longitudes_deg=conus_longitudes_deg,
query_latitudes_deg=query_latitudes_deg,
query_longitudes_deg=query_longitudes_deg,
use_shortcuts=True,
verbose=False)
if numpy.all(in_conus_flags):
return numpy.array([], dtype=int)
first_bad_index = numpy.where(numpy.invert(in_conus_flags))[0][0]
first_bad_longitude_deg = -1 * lng_conversion.convert_lng_negative_in_west(
query_longitudes_deg[first_bad_index])
print('Point ({0:.1f} deg N, {1:.1f} deg W) is not in CONUS!'.format(
query_latitudes_deg[first_bad_index], first_bad_longitude_deg))
object_not_in_conus_indices = numpy.unique(
query_object_indices[in_conus_flags == False])
bad_object_indices = numpy.array([], dtype=int)
for i in object_not_in_conus_indices:
these_indices = temporal_tracking.find_predecessors(
storm_object_table=storm_object_table,
target_row=i,
num_seconds_back=max_lead_time_sec,
max_num_sec_id_changes=1,
return_all_on_path=True)
bad_object_indices = numpy.concatenate(
(bad_object_indices, these_indices))
return numpy.unique(bad_object_indices)
