def test_save_to_msgpack():
data_file = os.path.join('simar', 'SIMAR_1052046')
file_name = 'full_simar_preprocessed.msg'
path = 'D:\\REPOSITORIO GIT\\protocol_project\\data\\intermediate_files'
# Read simar
data_simar, _ = read.simar(data_file, tests.full_data_path)
# Preproccesing simar
time_step = missing_values.find_timestep(data_simar, n_random_values=10)
data_clean = missing_values.erase_null_values(data_simar, method='all')
data_simar_interp = missing_values.fill_missing_values(
data_clean,
time_step,
technique='interpolation',
method='nearest',
limit=720,
limit_direction='both')
# Check missing values
miss_values = missing_values.find_missing_values(data_simar_interp,
time_step)
# Save simar
if miss_values.empty:
save_to_msgpack(data_simar_interp, file_name, path)
return
def test_extreme_events_full_simar_with_interpolation():
data_file = 'SIMAR_1052046'
full_data_path = os.path.join('..', '..', '..', '..', '..', 'data',
'simar')
data_simar, code = read.simar(data_file, path=full_data_path)
var_name = 'Hm0'
threshold = np.percentile(data_simar[var_name], 95)
minimum_interarrival_time = pd.Timedelta('3 days')
minimum_cycle_length = pd.Timedelta('3 hours')
cycles, calm_periods = extremal.extreme_events(data_simar,
var_name,
threshold,
minimum_interarrival_time,
minimum_cycle_length,
interpolation=True)
test = extremal.events_boundaries(cycles)
maximum = extremal.events_max(cycles)
# frequency = pd.Series(data.index).diff().min()
duration = extremal.events_duration(cycles)
magnitude = extremal.events_magnitude(cycles, threshold)
# noinspection PyTypeChecker
print(len(cycles) == len(calm_periods))
def test_min_cycles_duration():
# Read data
data_file = 'SIMAR_1052046'
full_data_path = os.path.join('..', '..', '..', '..', '..', 'data',
'simar')
data_simar, code = read.simar(data_file, path=full_data_path)
# Input
threshold = np.percentile(data_simar['Hm0'], 95)
minimum_interarrival_time = pd.Timedelta('3 days')
minimum_cycle_length = pd.Timedelta('3 hours')
# Cycles calculation
cycles, calm_periods = extremal.extreme_events(data_simar, 'Hm0',
threshold,
minimum_interarrival_time,
minimum_cycle_length)
# Cycles duration
cycles_duration = extremal.events_duration(cycles)
# Check if the min duration is equal to the set duration threshold
min_cycles_duration = cycles_duration.min()
# Find the cycles with duration less than the threshold
cont = 0
list_wrong_cycles = []
for cycle in cycles_duration:
if cycle == pd.Timedelta('2 hours'):
list_wrong_cycles.append(cont)
cont += 1
assert min_cycles_duration == minimum_cycle_length
def test_fill_missing_values():
# Read simar file
data_path = os.path.join(tests.sample_data_path, 'simar')
# noinspection PyTypeChecker
data_simar, _ = read.simar('SIMAR_1052046_short_gap', data_path)
# Calculate the time step
time_step = missing_values.find_timestep(data_simar)
# Fill missing values
data_fill = missing_values.fill_missing_values(data_simar,
time_step,
technique='interpolation',
method='nearest',
limit=24,
limit_direction='both')
tolerance = 0.01
assert data_fill.loc['1958-01-04 08',
'Hm0'] == pytest.approx(2.1, tolerance)
assert data_fill.loc['1958-01-04 12',
'Tp'] == pytest.approx(10.5, tolerance)
assert data_fill.loc['1958-01-04 14',
'Tp'] == pytest.approx(10.6, tolerance)
assert data_fill.loc['1960-12-31 19',
'Hm0'] == pytest.approx(0.6, tolerance)
def test_missing_values_report():
# Input
file_name = 'gaps_report.csv'
path = os.path.join('.', '..', '..', 'report', 'tests', 'output', 'tables')
# Read simar file
data_path = os.path.join(tests.sample_data_path, 'simar')
# noinspection PyTypeChecker
data_simar, _ = read.simar('SIMAR_1052046_short_gap', data_path)
# Calculate the time step
time_step = missing_values.find_timestep(data_simar)
# Find gaps
data_gaps = missing_values.find_missing_values(data_simar, time_step)
# Gaps report
data_gaps_report = missing_values.missing_values_report(
data_simar, data_gaps)
missing_values.missing_values_report_to_file(data_gaps_report, file_name,
path)
# Plot
missing_values.plot_missing_values(data=data_simar,
data_column='Hm0',
data_gaps=data_gaps,
title='',
var_name='Hm0',
var_unit='m',
fig_filename='',
circular=False,
label='Hm0')
def test_find_timestep():
# Read simar file
data_path = os.path.join(tests.sample_data_path, 'simar')
# noinspection PyTypeChecker
data_simar, _ = read.simar('SIMAR_1052046_short', data_path)
t_step = missing_values.find_timestep(data_simar)
assert t_step == pd.timedelta(hours=1)
def test_erase_null_values():
# Define null values
method = 'any'
# Read simar file
data_path = os.path.join(tests.sample_data_path, 'simar')
# noinspection PyTypeChecker
data_simar, _ = read.simar('SIMAR_1052046_short_gap', data_path)
# Erase null values
data = missing_values.erase_null_values(data_simar, method)
assert '1958-01-04 08' not in data.index
assert '1958-01-04 19' not in data.index
assert '1960-12-31 18' not in data.index
def test_find_missing_values():
# Read simar file
data_path = os.path.join(tests.sample_data_path, 'simar')
# noinspection PyTypeChecker
data_simar, _ = read.simar('SIMAR_1052046_short_gap', data_path)
# Calculate the time step
time_step = missing_values.find_timestep(data_simar)
data_gaps = missing_values.find_missing_values(data_simar, time_step)
# Representation of the gaps
fig = plt.figure()
ax = fig.gca()
ax.plot(data_simar.loc[:, 'Hm0'])
ax.plot(data_simar.loc[data_gaps.loc[:, 'pos_ini'], 'Hm0'],
'k.',
markersize=10)
ax.plot(data_simar.loc[data_gaps.loc[:, 'pos_fin'], 'Hm0'],
'k.',
markersize=10)
fig.show()
def test_gev_fit_to_annual_maxima_confidence_bands():
# Inputs
data_file = 'SIMAR_1052046'
threshold_percentile = 95
n_sim_boot = 100
alpha = 0.05 # Confidence level
# Read SIMAR
full_data_path = os.path.join('..', '..', '..', '..', '..', 'data',
'simar')
data_simar, code = read.simar(data_file, path=full_data_path)
threshold = np.percentile(data_simar.loc[:, 'Hm0'], threshold_percentile)
# Calculation of the annual maxima sample
annual_maxima = extremal.annual_maxima_calculation(data_simar['Hm0'])
# Annual Maxima Empirical distribution
ecdf_am = empirical_distributions.ecdf_histogram(annual_maxima)
ecdf_am_rp = extremal.return_period_curve(1, ecdf_am)
# Fit Annual Maxima to GEV
(param_orig, x_gev, y_gev, y_gev_rp) = extremal.extremal_distribution_fit(
data=data_simar,
var_name='Hm0',
sample=annual_maxima,
threshold=threshold,
fit_type='gev',
x_min=0.90 * min(annual_maxima),
x_max=1.5 * max(annual_maxima),
n_points=1000,
cumulative=True)
# Add confidence bands to asses the uncertainty (Bootstrapping)
boot_extreme = extremal.extremal_distribution_fit_bootstrapping(
sample=annual_maxima,
n_sim_boot=n_sim_boot,
data=data_simar,
var_name='Hm0',
threshold=threshold,
param_orig=param_orig,
fit_type='gev',
x_min=0.90 * min(annual_maxima),
x_max=1.5 * max(annual_maxima),
alpha=alpha)
# Representation
extremal.plot_extremal_cdf(x_gev,
y_gev,
ecdf_am,
n_sim_boot,
boot_extreme,
alpha,
title='',
var_name='Hm0',
var_unit='m',
fig_filename='',
circular=False,
extremal_label='GEV Fit',
empirical_label='GEV ECDF')
extremal.plot_extremal_return_period(x_gev,
y_gev_rp,
ecdf_am_rp,
n_sim_boot,
boot_extreme,
alpha,
title='',
var_name='Hm0',
var_unit='m',
fig_filename='',
circular=False,
extremal_label='GEV Fit',
empirical_label='GEV ECDF')
def test_gpd_fit_to_pot_confidence_bands():
# Inputs
data_file = 'SIMAR_1052046'
threshold_percentile = 95
minimum_interarrival_time = pd.Timedelta('3 days')
minimum_cycle_length = pd.Timedelta('3 hours')
interpolation = True
interpolation_method = 'linear'
interpolation_freq = '1min'
truncate = False
extra_info = False
n_sim_boot = 100
alpha = 0.05 # Confidence level
# Read SIMAR
full_data_path = os.path.join('..', '..', '..', '..', '..', 'data',
'simar')
data_simar, code = read.simar(data_file, path=full_data_path)
threshold = np.percentile(data_simar.loc[:, 'Hm0'], threshold_percentile)
# Storm cycles calculation
cycles, calm_periods = extremal.extreme_events(
data_simar, 'Hm0', threshold, minimum_interarrival_time,
minimum_cycle_length, interpolation, interpolation_method,
interpolation_freq, truncate, extra_info)
# Peaks over threshold
peaks_over_thres = extremal.events_max(cycles)
# POT Empirical distribution
ecdf_pot = empirical_distributions.ecdf_histogram(peaks_over_thres)
n_peaks_year = len(peaks_over_thres) / len(
data_simar['Hm0'].index.year.unique())
ecdf_pot_rp = extremal.return_period_curve(n_peaks_year, ecdf_pot)
# Fit POT to Scipy-GPD
(param_orig, x_gpd, y_gpd, y_gpd_rp) = extremal.extremal_distribution_fit(
data=data_simar,
var_name='Hm0',
sample=peaks_over_thres,
threshold=threshold,
fit_type='gpd',
x_min=0.90 * min(peaks_over_thres),
x_max=1.5 * max(peaks_over_thres),
n_points=1000,
cumulative=True)
# Add confidence bands to asses the uncertainty (Bootstrapping)
boot_extreme = extremal.extremal_distribution_fit_bootstrapping(
sample=peaks_over_thres,
n_sim_boot=n_sim_boot,
data=data_simar,
var_name='Hm0',
threshold=threshold,
param_orig=param_orig,
fit_type='gpd',
x_min=0.90 * min(peaks_over_thres),
x_max=1.5 * max(peaks_over_thres),
alpha=alpha)
# Representation
extremal.plot_extremal_cdf(x_gpd,
y_gpd,
ecdf_pot,
n_sim_boot,
boot_extreme,
alpha,
title='',
var_name='Hm0',
var_unit='m',
fig_filename='',
circular=False,
extremal_label='GPD Fit',
empirical_label='POT ECDF')
extremal.plot_extremal_return_period(x_gpd,
y_gpd_rp,
ecdf_pot_rp,
n_sim_boot,
boot_extreme,
alpha,
title='',
var_name='Hm0',
var_unit='m',
fig_filename='',
circular=False,
extremal_label='GPD Fit',
empirical_label='POT ECDF')
def test_poisson_pareto_fit_to_pot_and_gev_fit_to_annual_maxima():
# Inputs
data_file = 'SIMAR_1052046'
threshold_percentile = 95
minimum_interarrival_time = pd.Timedelta('3 days')
minimum_cycle_length = pd.Timedelta('3 hours')
interpolation = True
interpolation_method = 'linear'
interpolation_freq = '1min'
truncate = False
extra_info = False
# Read SIMAR
full_data_path = os.path.join('..', '..', '..', '..', '..', 'data',
'simar')
data_simar, code = read.simar(data_file, path=full_data_path)
threshold = np.percentile(data_simar.loc[:, 'Hm0'], threshold_percentile)
# Storm cycles calculation
cycles, calm_periods = extremal.extreme_events(
data_simar, 'Hm0', threshold, minimum_interarrival_time,
minimum_cycle_length, interpolation, interpolation_method,
interpolation_freq, truncate, extra_info)
# Peaks over threshold
peaks_over_thres = extremal.events_max(cycles)
# Calculation of the annual maxima sample
annual_maxima = extremal.annual_maxima_calculation(data_simar['Hm0'])
# POT Empirical distribution
ecdf_pot = empirical_distributions.ecdf_histogram(peaks_over_thres)
n_peaks_year = len(peaks_over_thres) / len(
data_simar['Hm0'].index.year.unique())
ecdf_pot_rp = extremal.return_period_curve(n_peaks_year, ecdf_pot)
# Annual Maxima Empirical distribution
ecdf_am = empirical_distributions.ecdf_histogram(annual_maxima)
ecdf_am_rp = extremal.return_period_curve(1, ecdf_am)
# Fit Annual Maxima to GEV
(param, x_gev, y_gev, y_gev_rp) = extremal.extremal_distribution_fit(
data=data_simar,
var_name='Hm0',
sample=annual_maxima,
threshold=None,
fit_type='gev',
x_min=0.90 * min(annual_maxima),
x_max=1.5 * max(annual_maxima),
n_points=1000,
cumulative=True)
# Fit Peaks over threshold to Poisson Pareto
(param, x_pp, y_pp, y_pp_rp) = extremal.extremal_distribution_fit(
data=data_simar,
var_name='Hm0',
sample=peaks_over_thres,
threshold=threshold,
fit_type='poisson',
x_min=0.90 * min(peaks_over_thres),
x_max=1.5 * max(peaks_over_thres),
n_points=1000,
cumulative=True)
# Represent results
plt.figure()
ax = plt.axes()
ax.plot(ecdf_am.index, ecdf_am, '.k', label='Annual maxima ECDF')
ax.plot(x_gev, y_gev, 'k', label='GEV fit')
ax.plot(x_pp, y_pp, label='Poisson-Pareto fit')
plt.xlabel('Hm0 (m)')
plt.ylabel('CDF')
ax.legend()
plt.grid()
plt.show()
plt.figure()
ax = plt.axes()
ax.semilogx(ecdf_am_rp, ecdf_am_rp.index, '.k', label='Annual maxima ECDF')
ax.semilogx(y_gev_rp, x_gev, 'k', label='GEV fit')
ax.semilogx(y_pp_rp, x_pp, label='Poisson-Pareto fit')
plt.xlim(0, 500)
plt.xlabel('Return Period (years)')
plt.ylabel('Hm0 (m)')
ax.legend()
plt.grid()
plt.show()
def read_full_simar(data_file=full_data_file, data_path=full_data_path):
# noinspection PyTypeChecker
data_simar, code = read.simar(data_file, data_path)
return data_simar
def read_sample_simar(data_file=sample_data_file, data_path=sample_data_path):
# noinspection PyTypeChecker
data_simar, code = read.simar(data_file, data_path)
return data_simar
def test_simar_data():
data_simar, _ = read.simar(tests.sample_data_file, tests.sample_data_path)
assert data_simar.shape[0] == 216
assert data_simar.ix[153, 'Tp'] == 11.2
def test_simar_code():
_, code = read.simar(tests.sample_data_file, tests.sample_data_path)
assert code == '1052046'
