def test_mpi_he_plot_bootstrapped_skill_over_leadyear_ds(
PM_ds_initialized_1d, PM_ds_control_1d):
"""
Checks plots from bootstrap_perfect_model works for xr.Dataset with one variable.
"""
res = bootstrap_perfect_model(
PM_ds_initialized_1d,
PM_ds_control_1d,
metric="pearson_r",
iterations=ITERATIONS,
)
res_ax = plot_bootstrapped_skill_over_leadyear(res)
assert res_ax is not None
def test_mpi_pm_plot_bootstrapped_skill_over_leadyear_da(
PM_da_initialized_1d, PM_da_control_1d):
"""
Checks plots from bootstrap_perfect_model works for xr.DataArray.
"""
res = bootstrap_perfect_model(
PM_da_initialized_1d,
PM_da_control_1d,
metric='pearson_r',
iterations=ITERATIONS,
)
res_ax = plot_bootstrapped_skill_over_leadyear(res)
assert res_ax is not None
def test_PerfectModelEnsemble_plot_bootstrapped_skill_over_leadyear(
perfectModelEnsemble_initialized_control, ):
"""
Checks plots from PerfectModelEnsemble.bootstrap().
"""
res = perfectModelEnsemble_initialized_control.bootstrap(
metric="pearson_r",
iterations=ITERATIONS,
reference=["uninitialized", "persistence"],
comparison="m2e",
dim=["init", "member"],
)
res_ax = plot_bootstrapped_skill_over_leadyear(res)
assert res_ax is not None
def test_mpi_he_plot_bootstrapped_skill_over_leadyear_single_uninit_lead(
PM_da_initialized_1d, PM_da_control_1d):
"""
Checks plots from bootstrap_perfect_model works for xr.DataArray.
"""
res = bootstrap_perfect_model(
PM_da_initialized_1d,
PM_da_control_1d,
metric="pearson_r",
iterations=ITERATIONS,
)
# set all but first uninit lead to nan
res[:, 2, 1:] = [np.nan] * (res.lead.size - 1)
res_ax = plot_bootstrapped_skill_over_leadyear(res)
assert res_ax is not None
def test_mpi_pm_plot_bootstrapped_skill_over_leadyear():
"""
Checks plots from bootstrap_perfect_model works.
"""
da = load_dataset('MPI-PM-DP-1D').isel(area=1, period=-1)
PM_da_ds1d = da['tos']
da = load_dataset('MPI-control-1D').isel(area=1, period=-1)
PM_da_control1d = da['tos']
# sig = 95
bootstrap = 5
res = bootstrap_perfect_model(
PM_da_ds1d, PM_da_control1d, metric='pearson_r', bootstrap=bootstrap
)
res_ax = plot_bootstrapped_skill_over_leadyear(res, 95)
assert res_ax is not None
axes[1, 0].legend(fontsize=10, frameon=False)
plt.tight_layout(h_pad=.1)
mpl.rcParams['savefig.dpi'] = 300
mpl.rcParams['savefig.format'] = 'jpg'
# plot and save
plot_fig2(bs_acc, bs_rmse)
if savefig:
plt.savefig(path_paper + 'Figure1_predictability_skill',
bbox_inches='tight')
# extra plot for OSM 2020
savefig = True
fig, ax = plt.subplots(1, 3, figsize=(18, 3))
plot_bootstrapped_skill_over_leadyear(bs_acc['CO2'].isel(
lead=slice(None, 7)), 95, plot_persistence=False, ax=ax[0])
ax[0].set_ylabel('ACC [ ]')
# ax[0].get_legend().remove()
plot_bootstrapped_skill_over_leadyear(bs_rmse['CO2'].isel(
lead=slice(None, 7)), 95, plot_persistence=False, ax=ax[1])
ax[1].set_ylabel('RMSE [ppm]')
ax[1].get_legend().remove()
plot_both_skill(bs_acc['CO2'], bs_rmse['CO2'],
ax=ax[2], unit='ppm', not_all=True, sig=95)
ax[2].legend(loc='lower left', frameon=False)
ax[2].set_xlabel('RMSE [ppm]')
if savefig:
plt.savefig(path_paper + 'Figure_predictability_skill',
bbox_inches='tight')