def test_real_case():
""""""
# inputs
precip, metadata = get_precipitation_fields(
num_prev_files=2,
num_next_files=0,
return_raw=False,
metadata=True,
upscale=2000,
)
# motion
motion = dense_lucaskanade(precip)
# parameters
timesteps = [1, 2, 3]
thr = 1 # mm / h
slope = 1 * metadata["accutime"] # min-1
# compute probability forecast
extrap_kwargs = dict(allow_nonfinite_values=True)
fct = forecast(precip[-1],
motion,
timesteps,
thr,
slope=slope,
extrap_kwargs=extrap_kwargs)
assert fct.ndim == 3
assert fct.shape[0] == len(timesteps)
assert fct.shape[1:] == precip.shape[1:]
assert np.nanmax(fct) <= 1.0
assert np.nanmin(fct) >= 0.0
# read precip field
date = datetime.strptime("201607112100", "%Y%m%d%H%M")
fns = io.find_by_date(date,
root,
fmt,
pattern,
ext,
timestep,
num_prev_files=2)
importer = io.get_method(importer_name, "importer")
precip, __, metadata = io.read_timeseries(fns, importer, **importer_kwargs)
precip, metadata = utils.to_rainrate(precip, metadata)
# precip[np.isnan(precip)] = 0
# motion
motion = dense_lucaskanade(precip)
# parameters
nleadtimes = 6
thr = 1 # mm / h
slope = 1 * timestep # km / min
# compute probability forecast
extrap_kwargs = dict(allow_nonfinite_values=True)
fct = forecast(precip[-1],
motion,
nleadtimes,
thr,
slope=slope,
extrap_kwargs=extrap_kwargs)
# Nicely print the metadata
pprint(metadata)
###############################################################################
# Deterministic nowcast with S-PROG
# ---------------------------------
#
# First, the motiong field is estimated using a local tracking approach based
# on the Lucas-Kanade optical flow.
# The motion field can then be used to generate a deterministic nowcast with
# the S-PROG model, which implements a scale filtering appraoch in order to
# progressively remove the unpredictable spatial scales during the forecast.
# Estimate the motion field
V = dense_lucaskanade(R)
# The S-PROG nowcast
nowcast_method = nowcasts.get_method("sprog")
R_f = nowcast_method(
R[-3:, :, :],
V,
n_leadtimes,
n_cascade_levels=6,
R_thr=-10.0,
)
# Back-transform to rain rate
R_f = transformation.dB_transform(R_f, threshold=-10.0, inverse=True)[0]
# Plot the S-PROG forecast
rainrate, metadata = conversion.to_rainrate(reflectivity, metadata)
# Upscale data to 2 km to reduce computation time
rainrate, metadata = dimension.aggregate_fields_space(rainrate, metadata, 2000)
# Plot the most recent rain rate field
plt.figure()
plot_precip_field(rainrate[-1, :, :])
plt.show()
###############################################################################
# Estimate the advection field
# ----------------------------
# The advection field is estimated using the Lucas-Kanade optical flow
advection = dense_lucaskanade(rainrate, verbose=True)
###############################################################################
# Deterministic nowcast
# ---------------------
# Compute 30-minute LINDA nowcast with 8 parallel workers
# Restrict the number of features to 15 to reduce computation time
nowcast_linda = linda.forecast(
rainrate,
advection,
6,
max_num_features=15,
add_perturbations=False,
num_workers=8,
measure_time=True,
