diff --git a/PE_CBL.py b/PE_CBL.py
index e37393f049d3d61ef810a91503c29ae4183e3794..f348c9ca395331fd027ffc0c57e258904141cadc 100644
--- a/PE_CBL.py
+++ b/PE_CBL.py
@@ -98,10 +98,10 @@ if __name__ == '__main__':
# Decide what figures to plot
fig01 = False
fig02 = True
- fig03 = False
- fig04 = False
- fig05 = False
- fig06 = False
+ fig03 = True
+ fig04 = True
+ fig05 = True
+ fig06 = True
fig07 = True
fig08 = True
@@ -110,9 +110,11 @@ if __name__ == '__main__':
opt02 = False # assimilation of profiles at two times
# Whether or not to run experiments without parameter estimation
+ # Applies only to sets B&C; no-PE run is always computed for experiments A&D
noPE_runs = False
- # Default PE experiment
+ ############################################################################
+ # Experiment A (control)
# Create a copy of the default settings
cbl_settings_A = dict(default_cbl_settings)
da_settings_A = dict(default_da_settings)
@@ -148,6 +150,279 @@ if __name__ == '__main__':
exp_A_noPE = experiment(da_settings_A_noPE)
pickle.dump(exp_A_noPE, open('exp_A_noPE.pickle', 'wb'))
+ ########################################################################
+ # Experiment B1
+ # Create a copy of the default settings
+ # Then re-use the available nature run and derived information
+ cbl_settings_B1 = dict(default_cbl_settings)
+ da_settings_B1 = dict(default_da_settings)
+ da_settings_B1['nr'] = deepcopy(nature_run)
+ da_settings_B1['truths'] = truths
+ da_settings_B1['observations'] = observations
+
+ # Change settings as necessary
+ cbl_settings_B1['perturbations_theta_amplitude'] = sigma_b_init*10
+ cbl_settings_B1['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
+ da_settings_B1['cbl_settings'] = cbl_settings_B1
+ da_settings_B1['obs_error_sdev_assimilate'] = np.ones(nobs)*sigma_o_as
+
+ # Run and save to disk
+ try:
+ exp_B1 = pickle.load(open("exp_B1.pickle", "rb"))
+ except:
+ exp_B1 = experiment(da_settings_B1)
+ setattr(exp_B1,'label','B1')
+ pickle.dump(exp_B1, open('exp_B1.pickle', 'wb'))
+
+ if noPE_runs:
+ # Corresponding experiment without parameter estimation
+ cbl_settings_B1_noPE = dict(cbl_settings_B1)
+ da_settings_B1_noPE = dict(da_settings_B1)
+ cbl_settings_B1_noPE['do_parameter_estimation'] = False
+ da_settings_B1_noPE['nr'] = deepcopy(nature_run)
+ da_settings_B1_noPE['nr'].do_parameter_estimation = False
+ da_settings_B1_noPE['cbl_settings'] = cbl_settings_B1_noPE
+
+ # Run and save to disk
+ try:
+ exp_B1_noPE = pickle.load(open("exp_B1_noPE.pickle", "rb"))
+ except:
+ exp_B1_noPE = experiment(da_settings_B1_noPE)
+ pickle.dump(exp_B1_noPE, open('exp_B1_noPE.pickle', 'wb'))
+
+ ########################################################################
+ # Experiment B2
+ # Create a copy of the default settings
+ # Then re-use the available nature run and derived information
+ cbl_settings_B2 = dict(default_cbl_settings)
+ da_settings_B2 = dict(default_da_settings)
+ da_settings_B2['nr'] = deepcopy(nature_run)
+ da_settings_B2['truths'] = truths
+ da_settings_B2['observations'] = observations
+
+ # Change settings as necessary
+ cbl_settings_B2['perturbations_theta_amplitude'] = sigma_b_init
+ cbl_settings_B2['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
+ da_settings_B2['cbl_settings'] = cbl_settings_B2
+ da_settings_B2['obs_error_sdev_assimilate'] = np.ones(nobs)*sigma_o_as*10
+
+ # Run and save to disk
+ try:
+ exp_B2 = pickle.load(open("exp_B2.pickle", "rb"))
+ except:
+ exp_B2 = experiment(da_settings_B2)
+ setattr(exp_B2,'label','B2')
+ pickle.dump(exp_B2, open('exp_B2.pickle', 'wb'))
+
+ if noPE_runs:
+ # Corresponding experiment without parameter estimation
+ cbl_settings_B2_noPE = dict(cbl_settings_B2)
+ da_settings_B2_noPE = dict(da_settings_B2)
+ cbl_settings_B2_noPE['do_parameter_estimation'] = False
+ da_settings_B2_noPE['nr'] = deepcopy(nature_run)
+ da_settings_B2_noPE['nr'].do_parameter_estimation = False
+ da_settings_B2_noPE['cbl_settings'] = cbl_settings_B2_noPE
+
+ # Run and save to disk
+ try:
+ exp_B2_noPE = pickle.load(open("exp_B2_noPE.pickle", "rb"))
+ except:
+ exp_B2_noPE = experiment(da_settings_B2_noPE)
+ pickle.dump(exp_B2_noPE, open('exp_B2_noPE.pickle', 'wb'))
+
+ ########################################################################
+ # Experiment B3
+ # Create a copy of the default settings
+ # Then re-use the available nature run and derived information
+ cbl_settings_B3 = dict(default_cbl_settings)
+ da_settings_B3 = dict(default_da_settings)
+ da_settings_B3['nr'] = deepcopy(nature_run)
+ da_settings_B3['truths'] = truths
+ da_settings_B3['observations'] = observations
+
+ # Change settings as necessary
+ cbl_settings_B3['perturbations_theta_amplitude'] = sigma_b_init*10
+ cbl_settings_B3['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
+ da_settings_B3['cbl_settings'] = cbl_settings_B3
+ da_settings_B3['obs_error_sdev_assimilate'] = np.ones(nobs)*sigma_o_as*10
+
+ # Run and save to disk
+ try:
+ exp_B3 = pickle.load(open("exp_B3.pickle", "rb"))
+ except:
+ exp_B3 = experiment(da_settings_B3)
+ setattr(exp_B3,'label','B3')
+ pickle.dump(exp_B3, open('exp_B3.pickle', 'wb'))
+
+ if noPE_runs:
+ # Corresponding experiment without parameter estimation
+ cbl_settings_B3_noPE = dict(cbl_settings_B3)
+ da_settings_B3_noPE = dict(da_settings_B3)
+ cbl_settings_B3_noPE['do_parameter_estimation'] = False
+ da_settings_B3_noPE['nr'] = deepcopy(nature_run)
+ da_settings_B3_noPE['nr'].do_parameter_estimation = False
+ da_settings_B3_noPE['cbl_settings'] = cbl_settings_B3_noPE
+
+ # Run and save to disk
+ try:
+ exp_B3_noPE = pickle.load(open("exp_B3_noPE.pickle", "rb"))
+ except:
+ exp_B3_noPE = experiment(da_settings_B3_noPE)
+ pickle.dump(exp_B3_noPE, open('exp_B3_noPE.pickle', 'wb'))
+
+ ########################################################################
+ # Experiment B4
+ # Create a copy of the default settings
+ cbl_settings_B4 = dict(default_cbl_settings)
+ da_settings_B4 = dict(default_da_settings)
+
+ # Change settings as necessary
+ # Changes include generation of observations, so the existing nature run
+ # can't be reused.
+ cbl_settings_B4['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
+ da_settings_B4['cbl_settings'] = cbl_settings_B4
+ da_settings_B4['obs_error_sdev_generate'] = np.ones(nobs)*sigma_o_as*5
+ da_settings_B4['obs_error_sdev_assimilate'] = np.ones(nobs)*sigma_o_as*10
+
+ # Run and save to disk
+ try:
+ exp_B4 = pickle.load(open("exp_B4.pickle", "rb"))
+ except:
+ exp_B4 = experiment(da_settings_B4)
+ setattr(exp_B4,'label','B4')
+ pickle.dump(exp_B4, open('exp_B4.pickle', 'wb'))
+
+ if noPE_runs:
+ # Corresponding experiment without parameter estimation
+ cbl_settings_B4_noPE = dict(cbl_settings_B4)
+ da_settings_B4_noPE = dict(da_settings_B4)
+ cbl_settings_B4_noPE['do_parameter_estimation'] = False
+ da_settings_B4_noPE['cbl_settings'] = cbl_settings_B4_noPE
+
+ # Run and save to disk
+ try:
+ exp_B4_noPE = pickle.load(open("exp_B4_noPE.pickle", "rb"))
+ except:
+ exp_B4_noPE = experiment(da_settings_B4_noPE)
+ pickle.dump(exp_B4_noPE, open('exp_B4_noPE.pickle', 'wb'))
+
+ ########################################################################
+ # Experiment C1
+ # Create a copy of the default settings
+ cbl_settings_C1 = dict(default_cbl_settings)
+ da_settings_C1 = dict(default_da_settings)
+
+ # Re-use the available nature run and derived information
+ da_settings_C1['nr'] = deepcopy(nature_run)
+ da_settings_C1['truths'] = truths
+ da_settings_C1['observations'] = observations
+
+ # Change settings as necessary
+ cbl_settings_C1['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
+ cbl_settings_C1['is_cgrad'] = False
+ da_settings_C1['cbl_settings'] = cbl_settings_C1
+
+ # Run and save to disk
+ try:
+ exp_C1 = pickle.load(open("exp_C1.pickle", "rb"))
+ except:
+ exp_C1 = experiment(da_settings_C1)
+ setattr(exp_C1,'label','C1')
+ pickle.dump(exp_C1, open('exp_C1.pickle', 'wb'))
+
+ if noPE_runs:
+ # Corresponding experiment without parameter estimation
+ cbl_settings_C1_noPE = dict(cbl_settings_C1)
+ da_settings_C1_noPE = dict(da_settings_C1)
+ cbl_settings_C1_noPE['do_parameter_estimation'] = False
+ da_settings_C1_noPE['cbl_settings'] = cbl_settings_C1_noPE
+
+ # Run and save to disk
+ try:
+ exp_C1_noPE = pickle.load(open("exp_C1_noPE.pickle", "rb"))
+ except:
+ exp_C1_noPE = experiment(da_settings_C1_noPE)
+ pickle.dump(exp_C1_noPE, open('exp_C1_noPE.pickle', 'wb'))
+
+ ########################################################################
+ # Experiment C2
+ # Create a copy of the default settings
+ cbl_settings_C2 = dict(default_cbl_settings)
+ da_settings_C2 = dict(default_da_settings)
+
+ # Re-use the available nature run and derived information
+ da_settings_C2['nr'] = deepcopy(nature_run)
+ da_settings_C2['truths'] = truths
+ da_settings_C2['observations'] = observations
+
+ # Change settings as necessary
+ cbl_settings_C2['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
+ cbl_settings_C2['Hmax'] = 0.15
+ da_settings_C2['cbl_settings'] = cbl_settings_C2
+
+ # Run and save to disk
+ try:
+ exp_C2 = pickle.load(open("exp_C2.pickle", "rb"))
+ except:
+ exp_C2 = experiment(da_settings_C2)
+ setattr(exp_C2,'label','C2')
+ pickle.dump(exp_C2, open('exp_C2.pickle', 'wb'))
+
+ if noPE_runs:
+ # Corresponding experiment without parameter estimation
+ cbl_settings_C2_noPE = dict(cbl_settings_C2)
+ da_settings_C2_noPE = dict(da_settings_C2)
+ cbl_settings_C2_noPE['do_parameter_estimation'] = False
+ da_settings_C2_noPE['cbl_settings'] = cbl_settings_C2_noPE
+
+ # Run and save to disk
+ try:
+ exp_C2_noPE = pickle.load(open("exp_C2_noPE.pickle", "rb"))
+ except:
+ exp_C2_noPE = experiment(da_settings_C2_noPE)
+ pickle.dump(exp_C2_noPE, open('exp_C2_noPE.pickle', 'wb'))
+
+ ########################################################################
+ # Experiment D
+ # Create a copy of the default settings
+ cbl_settings_D = dict(default_cbl_settings)
+ da_settings_D = dict(default_da_settings)
+
+ # Change settings as necessary
+ # Changes include generation of observations, so the existing nature run
+ # can't be reused.
+ cbl_settings_D['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
+ cbl_settings_D['perturbations_theta_amplitude'] = sigma_b_init*10
+ cbl_settings_D['Hmax'] = 0.15
+ cbl_settings_D['is_cgrad'] = False
+ cbl_settings_D['simulate_error_growth'] = True
+ cbl_settings_D['error_growth_perturbations_amplitude'] = sigma_b_init*5
+ da_settings_D['cbl_settings'] = cbl_settings_D
+ da_settings_D['obs_error_sdev_generate'] = np.ones(nobs)*sigma_o_as*5
+ da_settings_D['obs_error_sdev_assimilate'] = np.ones(nobs)*sigma_o_as*10
+
+ # Run and save to disk
+ try:
+ exp_D = pickle.load(open("exp_D.pickle", "rb"))
+ except:
+ exp_D = experiment(da_settings_D)
+ setattr(exp_D,'label','D')
+ pickle.dump(exp_D, open('exp_D.pickle', 'wb'))
+
+ # Experiment matching D, but without parameter estimation
+ cbl_settings_D_noPE = dict(cbl_settings_D)
+ da_settings_D_noPE = dict(da_settings_D)
+ cbl_settings_D_noPE['do_parameter_estimation'] = False
+ da_settings_D_noPE['cbl_settings'] = cbl_settings_D_noPE
+
+ try:
+ exp_D_noPE = pickle.load(open("exp_D_noPE.pickle", "rb"))
+ except:
+ exp_D_noPE = experiment(da_settings_D_noPE)
+ pickle.dump(exp_A_noPE, open('exp_D_noPE.pickle', 'wb'))
+
+ ########################################################################
+
if fig01:
# Create a copy of the default settings
@@ -234,7 +509,6 @@ if __name__ == '__main__':
if fig02:
- # Make plots
fig, [[ax0, ax1], [ax2, ax3]] = p.subplots(2,2,constrained_layout=True)
fig.set_size_inches(6,6)
#
@@ -272,137 +546,6 @@ if __name__ == '__main__':
if fig04:
- ########################################################################
- # Create a copy of the default settings
- # Then re-use the available nature run and derived information
- cbl_settings_B1 = dict(default_cbl_settings)
- da_settings_B1 = dict(default_da_settings)
- da_settings_B1['nr'] = deepcopy(nature_run)
- da_settings_B1['truths'] = truths
- da_settings_B1['observations'] = observations
-
- # Change settings as necessary
- cbl_settings_B1['perturbations_theta_amplitude'] = sigma_b_init*10
- cbl_settings_B1['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
- da_settings_B1['cbl_settings'] = cbl_settings_B1
- da_settings_B1['obs_error_sdev_assimilate'] = np.ones(nobs)*sigma_o_as
-
- # Run and save to disk
- exp_B1 = experiment(da_settings_B1)
- setattr(exp_B1,'label','B1')
- pickle.dump(exp_B1, open('exp_B1.pickle', 'wb'))
-
- if noPE_runs:
- # Corresponding experiment without parameter estimation
- cbl_settings_B1_noPE = dict(cbl_settings_B1)
- da_settings_B1_noPE = dict(da_settings_B1)
- cbl_settings_B1_noPE['do_parameter_estimation'] = False
- da_settings_B1_noPE['nr'] = deepcopy(nature_run)
- da_settings_B1_noPE['nr'].do_parameter_estimation = False
- da_settings_B1_noPE['cbl_settings'] = cbl_settings_B1_noPE
-
- # Run and save to disk
- exp_B1_noPE = experiment(da_settings_B1_noPE)
- pickle.dump(exp_B1_noPE, open('exp_B1_noPE.pickle', 'wb'))
-
- ########################################################################
- # Create a copy of the default settings
- # Then re-use the available nature run and derived information
- cbl_settings_B2 = dict(default_cbl_settings)
- da_settings_B2 = dict(default_da_settings)
- da_settings_B2['nr'] = deepcopy(nature_run)
- da_settings_B2['truths'] = truths
- da_settings_B2['observations'] = observations
-
- # Change settings as necessary
- cbl_settings_B2['perturbations_theta_amplitude'] = sigma_b_init
- cbl_settings_B2['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
- da_settings_B2['cbl_settings'] = cbl_settings_B2
- da_settings_B2['obs_error_sdev_assimilate'] = np.ones(nobs)*sigma_o_as*10
-
- # Run and save to disk
- exp_B2 = experiment(da_settings_B2)
- setattr(exp_B2,'label','B2')
- pickle.dump(exp_B2, open('exp_B2.pickle', 'wb'))
-
- if noPE_runs:
- # Corresponding experiment without parameter estimation
- cbl_settings_B2_noPE = dict(cbl_settings_B2)
- da_settings_B2_noPE = dict(da_settings_B2)
- cbl_settings_B2_noPE['do_parameter_estimation'] = False
- da_settings_B2_noPE['nr'] = deepcopy(nature_run)
- da_settings_B2_noPE['nr'].do_parameter_estimation = False
- da_settings_B2_noPE['cbl_settings'] = cbl_settings_B2_noPE
-
- # Run and save to disk
- exp_B2_noPE = experiment(da_settings_B2_noPE)
- pickle.dump(exp_B2_noPE, open('exp_B2_noPE.pickle', 'wb'))
-
- ########################################################################
- # Create a copy of the default settings
- # Then re-use the available nature run and derived information
- cbl_settings_B3 = dict(default_cbl_settings)
- da_settings_B3 = dict(default_da_settings)
- da_settings_B3['nr'] = deepcopy(nature_run)
- da_settings_B3['truths'] = truths
- da_settings_B3['observations'] = observations
-
- # Change settings as necessary
- cbl_settings_B3['perturbations_theta_amplitude'] = sigma_b_init*10
- cbl_settings_B3['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
- da_settings_B3['cbl_settings'] = cbl_settings_B3
- da_settings_B3['obs_error_sdev_assimilate'] = np.ones(nobs)*sigma_o_as*10
-
- # Run and save to disk
- exp_B3 = experiment(da_settings_B3)
- setattr(exp_B3,'label','B3')
- pickle.dump(exp_B3, open('exp_B3.pickle', 'wb'))
-
- if noPE_runs:
- # Corresponding experiment without parameter estimation
- cbl_settings_B3_noPE = dict(cbl_settings_B3)
- da_settings_B3_noPE = dict(da_settings_B3)
- cbl_settings_B3_noPE['do_parameter_estimation'] = False
- da_settings_B3_noPE['nr'] = deepcopy(nature_run)
- da_settings_B3_noPE['nr'].do_parameter_estimation = False
- da_settings_B3_noPE['cbl_settings'] = cbl_settings_B3_noPE
-
- # Run and save to disk
- exp_B3_noPE = experiment(da_settings_B3_noPE)
- pickle.dump(exp_B3_noPE, open('exp_B3_noPE.pickle', 'wb'))
-
- ########################################################################
- # Create a copy of the default settings
- cbl_settings_B4 = dict(default_cbl_settings)
- da_settings_B4 = dict(default_da_settings)
-
- # Change settings as necessary
- # Changes include generation of observations, so the existing nature run
- # can't be reused.
- cbl_settings_B4['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
- da_settings_B4['cbl_settings'] = cbl_settings_B4
- da_settings_B4['obs_error_sdev_generate'] = np.ones(nobs)*sigma_o_as*5
- da_settings_B4['obs_error_sdev_assimilate'] = np.ones(nobs)*sigma_o_as*10
-
- # Run and save to disk
- exp_B4 = experiment(da_settings_B4)
- setattr(exp_B4,'label','B4')
- pickle.dump(exp_B4, open('exp_B4.pickle', 'wb'))
-
- if noPE_runs:
- # Corresponding experiment without parameter estimation
- cbl_settings_B4_noPE = dict(cbl_settings_B4)
- da_settings_B4_noPE = dict(da_settings_B4)
- cbl_settings_B4_noPE['do_parameter_estimation'] = False
- da_settings_B4_noPE['cbl_settings'] = cbl_settings_B4_noPE
-
- # Run and save to disk
- exp_B4_noPE = experiment(da_settings_B4_noPE)
- pickle.dump(exp_B4_noPE, open('exp_B4_noPE.pickle', 'wb'))
-
- ########################################################################
-
- # Make plots
fig, [[ax1, ax2], [ax3, ax4]] = p.subplots(2,2,constrained_layout=True)
fig.set_size_inches(6,4)
#
@@ -436,70 +579,6 @@ if __name__ == '__main__':
if fig05:
- ########################################################################
- # Create a copy of the default settings
- cbl_settings_C1 = dict(default_cbl_settings)
- da_settings_C1 = dict(default_da_settings)
-
- # Re-use the available nature run and derived information
- da_settings_C1['nr'] = deepcopy(nature_run)
- da_settings_C1['truths'] = truths
- da_settings_C1['observations'] = observations
-
- # Change settings as necessary
- cbl_settings_C1['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
- cbl_settings_C1['is_cgrad'] = False
- da_settings_C1['cbl_settings'] = cbl_settings_C1
-
- # Run and save to disk
- exp_C1 = experiment(da_settings_C1)
- setattr(exp_C1,'label','C1')
- pickle.dump(exp_C1, open('exp_C1.pickle', 'wb'))
-
- if noPE_runs:
- # Corresponding experiment without parameter estimation
- cbl_settings_C1_noPE = dict(cbl_settings_C1)
- da_settings_C1_noPE = dict(da_settings_C1)
- cbl_settings_C1_noPE['do_parameter_estimation'] = False
- da_settings_C1_noPE['cbl_settings'] = cbl_settings_C1_noPE
-
- # Run and save to disk
- exp_C1_noPE = experiment(da_settings_C1_noPE)
- pickle.dump(exp_C1_noPE, open('exp_C1_noPE.pickle', 'wb'))
-
- ########################################################################
- # Create a copy of the default settings
- cbl_settings_C2 = dict(default_cbl_settings)
- da_settings_C2 = dict(default_da_settings)
-
- # Re-use the available nature run and derived information
- da_settings_C2['nr'] = deepcopy(nature_run)
- da_settings_C2['truths'] = truths
- da_settings_C2['observations'] = observations
-
- # Change settings as necessary
- cbl_settings_C2['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
- cbl_settings_C2['Hmax'] = 0.15
- da_settings_C2['cbl_settings'] = cbl_settings_C2
-
- # Run and save to disk
- exp_C2 = experiment(da_settings_C2)
- setattr(exp_C2,'label','C2')
- pickle.dump(exp_C2, open('exp_C2.pickle', 'wb'))
-
- if noPE_runs:
- # Corresponding experiment without parameter estimation
- cbl_settings_C2_noPE = dict(cbl_settings_C2)
- da_settings_C2_noPE = dict(da_settings_C2)
- cbl_settings_C2_noPE['do_parameter_estimation'] = False
- da_settings_C2_noPE['cbl_settings'] = cbl_settings_C2_noPE
-
- # Run and save to disk
- exp_C2_noPE = experiment(da_settings_C2_noPE)
- pickle.dump(exp_C2_noPE, open('exp_C2_noPE.pickle', 'wb'))
-
- ########################################################################
- # Make plots
fig, [ax1, ax2] = p.subplots(1,2,constrained_layout=True)
fig.set_size_inches(6,2)
#
@@ -550,29 +629,6 @@ if __name__ == '__main__':
if fig07:
- # Create a copy of the default settings
- cbl_settings_D = dict(default_cbl_settings)
- da_settings_D = dict(default_da_settings)
-
- # Change settings as necessary
- # Changes include generation of observations, so the existing nature run
- # can't be reused.
- cbl_settings_D['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
- cbl_settings_D['perturbations_theta_amplitude'] = sigma_b_init*10
- cbl_settings_D['Hmax'] = 0.15
- cbl_settings_D['is_cgrad'] = False
- cbl_settings_D['simulate_error_growth'] = True
- cbl_settings_D['error_growth_perturbations_amplitude'] = sigma_b_init*10
- da_settings_D['cbl_settings'] = cbl_settings_D
- da_settings_D['obs_error_sdev_generate'] = np.ones(nobs)*sigma_o_as*5
- da_settings_D['obs_error_sdev_assimilate'] = np.ones(nobs)*sigma_o_as*10
-
- # Run and save to disk
- exp_D = experiment(da_settings_D)
- setattr(exp_D,'label','D')
- pickle.dump(exp_D, open('exp_D.pickle', 'wb'))
-
- # Make plots
fig, [[ax0, ax1],[ax2,ax3]] = p.subplots(2,2,constrained_layout=True)
fig.set_size_inches(6,6)
#
@@ -600,41 +656,15 @@ if __name__ == '__main__':
if fig08:
- # Create a copy of the default settings
- cbl_settings_D = dict(default_cbl_settings)
- da_settings_D = dict(default_da_settings)
-
- # Change settings as necessary
- # Changes include generation of observations, so the existing nature run
- # can't be reused.
- cbl_settings_D['initial_perturbed_parameters'] = exp_A.da.initial_perturbed_parameters
- cbl_settings_D['perturbations_theta_amplitude'] = sigma_b_init*10
- cbl_settings_D['Hmax'] = 0.15
- cbl_settings_D['is_cgrad'] = False
- cbl_settings_D['simulate_error_growth'] = True
- cbl_settings_D['error_growth_perturbations_amplitude'] = sigma_b_init*10
- da_settings_D['cbl_settings'] = cbl_settings_D
- da_settings_D['obs_error_sdev_generate'] = np.ones(nobs)*sigma_o_as*10
- da_settings_D['obs_error_sdev_assimilate'] = np.ones(nobs)*sigma_o_as*10
-
- # Experiment matching D, but without parameter estimation
- cbl_settings_D_noPE = dict(cbl_settings_D)
- da_settings_D_noPE = dict(da_settings_D)
- cbl_settings_D_noPE['do_parameter_estimation'] = False
- da_settings_D_noPE['cbl_settings'] = cbl_settings_D_noPE
-
- exp_D = pickle.load(open("exp_D.pickle", "rb"))
- try:
- exp_D_noPE = pickle.load(open("exp_D_noPE.pickle", "rb"))
- except:
- exp_D_noPE = experiment(da_settings_D_noPE)
- pickle.dump(exp_A_noPE, open('exp_D_noPE.pickle', 'wb'))
+ experiments_pe = [exp_A]
+ experiments_nope = [exp_A_noPE]
+ labels = ["Exp. A"]
+ plot_diagnostics(experiments_pe,experiments_nope,labels,'fig08a.png')
experiments_pe = [exp_D]
experiments_nope = [exp_D_noPE]
labels = ["Exp. D"]
-
- plot_diagnostics(experiments_pe,experiments_nope,labels,'fig08.png')
+ plot_diagnostics(experiments_pe,experiments_nope,labels,'fig08d.png')
if opt01: