def plot_pred(date, time, fti, target, **kwargs):
"""Worker func for plotting ML predictions"""
examples = kwargs['examples']
info = kwargs['info']
time_marker = 'first_hour' if fti <= 12 else 'second_hour'
rng = np.arange(0., 0.55, 0.05) if 'torn' in target else np.arange(
0., 1.1, 0.1)
kwargs = {'alpha': 0.95, 'extend': 'neither', 'cmap': target_cmap[target]}
plt = Plotting(date=None,
z1_levels=rng,
z2_levels=[35, 60],
z3_levels=[35., 60],
z4_levels=[prob_threshold[target], 90.],
z4_color='k',
**kwargs)
n_panels = 2
fig, axes, = plt._create_fig(
fig_num=0,
sub_plots=(2, 1),
plot_map=False,
figsize=(7, 6),
hspace=0.15,
wspace=0.15,
)
map_axes, x, y = plt._generate_base_map(axes=axes, date=date, fig=fig)
valid_date_and_time, initial_date_and_time = get_time.determine_forecast_valid_datetime(
date_dir=str(date), time_dir=time, fcst_time_idx=fti)
try:
mrms_dbz = mrms.load_single_mrms_time(
date_dir=date,
valid_datetime=valid_date_and_time,
var_name='DZ_CRESSMAN')
except:
return None
try:
ens_data = EnsembleData(date_dir=date,
time_dir=time,
base_path='summary_files')
data = ens_data.load(variables=['comp_dz'],
time_indexs=[fti],
tag='ENS')
except:
return None
wofs_dbz = data['comp_dz'].values[0]
wofs_dbz = prob_match_mean(var=wofs_dbz,
mean_var=np.mean(wofs_dbz, axis=0),
neighborhood=15)
init_time_str, valid_time_str = get_timestampes(initial_date_and_time,
valid_date_and_time)
load_lsr = loadLSR(date_dir=date,
date=valid_date_and_time[0],
time=valid_date_and_time[1],
forecast_length=30)
load_wwa = loadWWA(date_dir=date,
date=valid_date_and_time[0],
time=valid_date_and_time[1],
forecast_length=30,
time_window=30)
hail_ll = load_lsr.load_hail_reports()
torn_ll = load_lsr.load_tornado_reports()
wind_ll = load_lsr.load_wind_reports()
lsr_points = {'hail': hail_ll, 'tornado': torn_ll, 'wind': wind_ll}
torn_wwa = load_wwa.load_tornado_warning_polygon(return_polygons=True)
wwa_points = {'tornado': torn_wwa}
ml_preds, objects = load_ml_predictions(time_marker,
target,
date,
time,
fti,
model_names=['LogisticRegression'],
examples=examples,
info=info)
baseline_pred = load_baseline(date, time, fti, target, time_marker)
ml_preds, objects = smooth_objects(objects,
ml_preds,
model_names=['LogisticRegression'])
if baseline_pred is None or ml_preds is None:
return None
all_preds = ml_preds + [baseline_pred]
titles = ['LogisticRegression', baseline_names[target]]
for j, forecast_probabilities in enumerate(all_preds):
ax = axes.flat[j]
if np.amax(forecast_probabilities.astype(float) *
100.) > prob_threshold[target]:
z4 = forecast_probabilities.astype(float) * 100.
else:
z4 = None
###wwa_points=wwa_points,
contours = plt.spatial_plotting(
fig,
ax,
x,
y,
lsr_points=lsr_points,
z1=np.ma.masked_where(forecast_probabilities == 0.,
forecast_probabilities),
z2=mrms_dbz,
z3=wofs_dbz,
z4=z4,
map_ax=map_axes[j],
)
ax.set_title(titles[j], fontsize=12, pad=1.0, alpha=0.8)
label_centroid(objects[0], forecast_probabilities, ax)
axes.flat[0].text(
0.0,
1.17,
f'Init Time : {init_time_str}',
fontsize=8,
alpha=0.9,
transform=axes.flat[0].transAxes,
)
axes.flat[0].text(0.0,
1.12,
f'Valid Time : {valid_time_str}',
fontsize=8,
alpha=0.9,
transform=axes.flat[0].transAxes)
base_plot.add_alphabet_label(n_panels, axes, pos=(0.9, 0.08))
additional_handles = [
Line2D([0], [0],
marker='o',
color='w',
markerfacecolor='r',
markersize=6,
alpha=0.8),
Line2D([0], [0],
marker='o',
color='w',
markerfacecolor='b',
markersize=6,
alpha=0.8),
Line2D([0], [0],
marker='o',
color='w',
markerfacecolor='g',
markersize=6,
alpha=0.8),
Line2D([0], [0], color='b', alpha=0.8),
Line2D([0], [0], color='k', alpha=0.8),
]
additional_labels = [
'Tornado', 'Severe Wind', 'Severe Hail', 'WoFS PMM DBZ > 35',
'MRMS DBZ > 35'
]
major_ax = base_plot.set_major_axis_labels(fig,
xlabel='',
ylabel_left='',
labelpad=5)
#base_plot.set_legend(n_panels, fig, axes[0,0], major_ax, additional_handles, additional_labels, bbox_to_anchor=(0.425, -0.375), ncol=2)
colorbar_labels = {
'tornado': 'Probability of Tornado',
'severe_hail': 'Probability of Severe Hail',
'severe_wind': 'Probability of Severe Wind'
}
base_plot.add_colorbar(fig=fig,
plot_obj=contours,
ax=axes,
colorbar_label=colorbar_labels[target])
fname = f'example_ml_predictions_vs_baseline_{date}_{time}_{target}_{time_marker}_{fti:02d}.png'
plt._save_fig(fig=fig, fname=fname, dpi=300, aformat='png')
date_dir=str(date), time_dir=time, fcst_time_idx=fti
)
mrms_dbz = mrms.load_single_mrms_time(date_dir=date, valid_datetime=initial_date_and_time, var_name='DZ_CRESSMAN' )
ens_data = EnsembleData(date_dir=date, time_dir=time, base_path='summary_files')
data = ens_data.load(variables=['comp_dz'], time_indexs=[0], tag='ENS')
wofs_dbz = data['comp_dz'].values[0]
wofs_dbz = prob_match_mean(var=wofs_dbz, mean_var=np.mean(wofs_dbz, axis=0), neighborhood=15)
init_time_str, valid_time_str = get_timestampes(initial_date_and_time, valid_date_and_time)
idx = get_examples_at_datetime(info, date, time, fti)
labeled_regions = load_probability_objects(date, time, fti)
forecast_probabilities = to_predictions_2d(predictions, labeled_regions, labels_column, idx)
map_axes, x, y = plt._generate_base_map(axes=[axes], date=date, fig=fig)
kwargs = {'alpha':0.95, 'extend': 'neither', 'cmap': target_cmap[target_var]}
plt = Plotting( date=None, z1_levels = np.arange(0., 1.1, .1), z2_levels = [35, 60], z3_levels=[35., 60], **kwargs )
contours = plt.spatial_plotting(fig,
axes,
x, y,
z1=np.ma.masked_where( forecast_probabilities==0., forecast_probabilities),
z2=mrms_dbz,
z3=wofs_dbz,
map_ax=map_axes[0],
)
if i %3 == 0:
contours_set.append(contours)