# 1.00000000e-16, 1.00000000e-16,
# 5.00000000e-14, 5.00000000e-16,
# 5.00000000e-14, 5.00000000e-16,
# 5.00000000e-14, 3.00000000e-14,
# 5.00000000e-15, 5.00000000e-15,
# 5.00000000e-14, 5.00000000e-16,
# 5.00000000e-14, 5.00000000e-16]))
#
# process_model_coarse = GC.GeoModel(name='test_process_model_kericini',
# datfile_name='input-files/kericini/coarse-model/Keriv0_027',
# incon_name='input-files/kericini/coarse-model/Keriv0_027',
# geom_name='input-files/kericini/coarse-model/gKerinci_v0',
# islayered=True)
bmodel = IC.BayesModel(name='test_bayes_model',
process_model=process_model_coarse)
#---test calculate model run time
start = timeit.default_timer()
bmodel.process_model.simulate(do_update_obs_wells=False)
stop = timeit.default_timer()
print('time to run model (s): ')
print(stop - start)
#sflat = pickle.load(
# open("./saved_data/sampler_flatchain_test_process_model_coarse.p", "rb"))
else:
synthetic_data = pickle.load(open("./saved_data/synthetic_data.p", "rb"))
synthetic_model_fine.ss_temps = synthetic_data['T_measured']
#syn_model.ss_temps = syn_model.T_measured
synthetic_model_fine.T_noise = synthetic_data['T_noise']
synthetic_model_fine.ss_temps_obs_well = synthetic_data['T_obs_well']
synthetic_model_fine.d_obs_well = synthetic_data['d_obs_well']
#synthetic_data = GC.GeoModel(name='test_synthetic_data',
# datfile_name='input-files/elvar-new-tighter-cap/fine-model/2DF002',
# incon_name='input-files/elvar-new-tighter-cap/fine-model/2DF002_IC',
# geom_name = 'input-files/elvar-new-tighter-cap/fine-model/g2fine',
# islayered=True)
#---create a parameter model
parameter_space = IC.ParameterSpace(mu=-15, sigma=1.5)
#----create a process space. Need for predictive checks.
process_space = IC.ProcessSpace()
#---create or load comparison space model (basis of likelihood function)
#pre-discrep.
measurement_space = IC.MeasurementSpace(bias=0.0, sigma=5.0) #without discrep.
#-----set up discrepancy info
discrepancy_filename = 'discrepancies_combined.p'
load_discrepancy = True
map_coarse_discrep_to_data_grid = False # - default should be False?
if load_discrepancy:
#---load real data of appropriate resolution and store in above.
real_data_model.d_obs_well = {}
real_data_model.ss_temps_obs_well = {}
for i,welli in enumerate(list_of_obs_wells):
df = pd.read_csv('./saved_data/kerinci_data/Temp_' + welli + '.dat',header=None,sep=' ')
df.rename(columns={1:'d',0:'T'},inplace=True)
real_data_model.d_obs_well[i] = df['d']
real_data_model.ss_temps_obs_well[i] = df['T']
#---create a basic comparison model (basis of likelihood function)
measurement_space = IC.MeasurementSpace(bias=0.0, sigma=5.0)
#---create a parameter model
parameter_space = IC.ParameterSpace(mu=-15, sigma=1.5)
#-----create a basic process space model
process_space = IC.ProcessSpace()
#---create a Bayes model
#use pro_model_coarse for coarse, pro_model_medium for medium
#bmodel = IC.BayesModel(name='test_bayes_model',
# process_model=process_model_medium,
# data_model=synthetic_model_fine,
# comparison_model=comparison_model,
# parameter_model=parameter_model)
real_data_model.d_obs_well = {}
real_data_model.ss_temps_obs_well = {}
for i, welli in enumerate(list_of_obs_wells):
df = pd.read_csv('./saved_data/kerinci_data/Temp_' + welli + '.dat',
header=None,
sep=' ')
df.rename(columns={1: 'd', 0: 'T'}, inplace=True)
real_data_model.d_obs_well[i] = df['d']
real_data_model.ss_temps_obs_well[i] = df['T']
#---create a basic comparison model (basis of likelihood function)
measurement_space = IC.MeasurementSpace(bias=0.0, sigma=10.0)
#---create a parameter model
parameter_space = IC.ParameterSpace(mu=-15, sigma=1.5)
#-----create a basic process space model
process_space = IC.ProcessSpace()
#-----set up discrepancy info
load_discrepancy = True
map_coarse_discrep_to_data_grid = False # - default should be False?
#discrepancy_filename = 'discrepancies_combined_kerinci.p'
discrepancy_filename = 'discrepancies_combined_kerinci_map_data.p'
if load_discrepancy:
discrep = IC.ModelDiscrep(process_space=process_space,
else:
synthetic_data = pickle.load(open("./saved_data/synthetic_data.p", "rb"))
synthetic_model_fine.ss_temps = synthetic_data['T_measured']
#syn_model.ss_temps = syn_model.T_measured
synthetic_model_fine.T_noise = synthetic_data['T_noise']
synthetic_model_fine.ss_temps_obs_well = synthetic_data['T_obs_well']
synthetic_model_fine.d_obs_well = synthetic_data['d_obs_well']
#synthetic_data = GC.GeoModel(name='test_synthetic_data',
# datfile_name='input-files/elvar-new-tighter-cap/fine-model/2DF002',
# incon_name='input-files/elvar-new-tighter-cap/fine-model/2DF002_IC',
# geom_name = 'input-files/elvar-new-tighter-cap/fine-model/g2fine',
# islayered=True)
#---create a basic comparison model (basis of likelihood function)
measurement_space = IC.MeasurementSpace(bias=0.0, sigma=5.0)
#---create a parameter model
parameter_space = IC.ParameterSpace(mu=-15, sigma=1.5)
#----create a process space. Need for predictive checks.
process_space = IC.ProcessSpace()
#use process_model_coarse for coarse, process_model_medium for medium.
bmodel = IC.BayesModel(name='test_bayes_model_med_fine_tighter_cap',
process_model=process_model_medium,
data_model=synthetic_model_fine,
measurement_space=measurement_space,
parameter_space=parameter_space,
process_space=process_space)
perm_powers=perm_powers_truths)
synthetic_data_model_fine.simulate()
if generate_new_data:
synthetic_data_model_fine.generate_synthetic_data(
perm_powers_truths=perm_powers_truths)
else:
synthetic_data = pickle.load(open("./saved_data/synthetic_data.p", "rb"))
synthetic_data_model_fine.ss_temps = synthetic_data['T_measured']
#syn_model.ss_temps = syn_model.T_measured
synthetic_data_model_fine.T_noise = synthetic_data['T_noise']
synthetic_data_model_fine.ss_temps_obs_well = synthetic_data['T_obs_well']
synthetic_data_model_fine.d_obs_well = synthetic_data['d_obs_well']
#---create a basic comparison model (basis of likelihood function)
measurement_space = IC.MeasurementSpace(bias=0.0, sigma=5.0)
#---create a parameter model
parameter_space = IC.ParameterSpace(mu=-15, sigma=1.5)
#-----create a basic process space model
process_space = IC.ProcessSpace()
#---create a Bayes model
#use pro_model_coarse for coarse, pro_model_medium for medium
#bmodel = IC.BayesModel(name='test_bayes_model',
# process_model=process_model_medium,
# data_model=synthetic_data_model_fine,
# comparison_model=comparison_model,
# parameter_model=parameter_model)
name='test_process_model_fine',
datfile_name='input-files/elvar-new-tighter-cap/fine-model/2DF002',
incon_name='input-files/elvar-new-tighter-cap/fine-model/2DF002_IC',
geom_name='input-files/elvar-new-tighter-cap/fine-model/g2fine',
islayered=True)
#---fine synthetic model
synthetic_model_fine = GC.GeoModel(
name='test_synthetic_model_fine',
datfile_name='input-files/elvar-new-tighter-cap/fine-model/2DF002',
incon_name='input-files/elvar-new-tighter-cap/fine-model/2DF002_IC',
geom_name='input-files/elvar-new-tighter-cap/fine-model/g2fine',
islayered=True)
#---create a basic comparison model (basis of likelihood function)
measurement_space = IC.MeasurementSpace(bias=0.0, sigma=5.0)
#just default/null process and parameter spaces required.
process_space = IC.ProcessSpace()
parameter_space = IC.ParameterSpace()
#sflat = pickle.load(
# open("./saved_data/sampler_flatchain_test_process_model_coarse.p", "rb"))
sflat = pickle.load(
open("./saved_data/sampler_flatchain_test_process_model_medium.p", "rb"))
param_sets = sflat
#param_sets = np.random.normal(loc=-15, scale=1.5, size=(30, 12))
#param_sets[:, 2:4] = -16