def test_random_mb_run():
rid = 'RGI60-15.03473'
gdir = oggm_compat.single_flowline_glacier_directory(rid, prepro_border=80)
mbmod = oggm_compat.RandomLinearMassBalance(gdir, seed=1, sigma_ela=300,
h_perc=55)
from oggm.core.flowline import robust_model_run
flmodel = robust_model_run(gdir, mb_model=mbmod, ys=0, ye=700)
# Check that "something" is computed
import xarray as xr
ds = xr.open_dataset(gdir.get_filepath('model_diagnostics'))
assert ds.isel(time=-1).volume_m3 > 0
if do_plot:
import matplotlib.pyplot as plt
from oggm import graphics
graphics.plot_modeloutput_section(flmodel)
f, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=(12, 4))
(ds.volume_m3 * 1e-9).plot(ax=ax1)
ax1.set_ylabel('Glacier volume (km$^{3}$)')
(ds.area_m2 * 1e-6).plot(ax=ax2)
ax2.set_ylabel('Glacier area (km$^{2}$)')
(ds.length_m * 1e3).plot(ax=ax3)
ax3.set_ylabel('Glacier length (km)')
plt.tight_layout()
plt.show()
def test_random_mb_run():
rid = 'RGI60-15.03473'
gdir = oggm_compat.single_flowline_glacier_directory(rid, prepro_border=80)
# This initializes the mass balance model, but does not run it
mbmod = oggm_compat.RandomLinearMassBalance(gdir,
seed=1,
sigma_ela=300,
h_perc=55)
# HERE CAN BE THE LOOP SUCH THAT EVERYTHING IS ALREADY LOADED
for i in [1, 2, 3, 4]:
# Change the model parameter
mbmod.param1 = i
# Run the mass balance model with fixed geometry
ts_mb = mbmod.get_specific_mb(years=[2000, 2001, 2002])
# Run the glacier flowline model with a mass balance model
from oggm.core.flowline import robust_model_run
flmodel = robust_model_run(gdir, mb_model=mbmod, ys=0, ye=700)
# Check that "something" is computed
import xarray as xr
ds = xr.open_dataset(gdir.get_filepath('model_diagnostics'))
assert ds.isel(time=-1).volume_m3 > 0
if do_plot:
import matplotlib.pyplot as plt
from oggm import graphics
graphics.plot_modeloutput_section(flmodel)
f, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=(12, 4))
(ds.volume_m3 * 1e-9).plot(ax=ax1)
ax1.set_ylabel('Glacier volume (km$^{3}$)')
(ds.area_m2 * 1e-6).plot(ax=ax2)
ax2.set_ylabel('Glacier area (km$^{2}$)')
(ds.length_m * 1e3).plot(ax=ax3)
ax3.set_ylabel('Glacier length (km)')
plt.tight_layout()
plt.show()
def run_uncertain_random_climate(gdir,
nyears=700,
output_filesuffix='',
sigma_t=None,
sigma_p=None,
sigma_smb=None,
rdn_temp_bias_seed=None,
rdn_prcp_bias_seed=None,
rdn_bias_seed=None,
orig_mb_tbias=0,
orig_mb_pbias=1,
orig_mb_sbias=0,
**kwargs):
"""Test stuff
Parameters
----------
gdir
nyears
output_filesuffix
rdn_temp_bias_seed
rdn_prcp_bias_seed
rdn_bias_seed
kwargs
Returns
-------
"""
# Find the optimal bias for a balanced MB
mbref = mbmods.PastMassBalance(gdir)
h, w = gdir.get_inversion_flowline_hw()
cyrs = np.unique(mbref.years)
def to_minimize(x):
mbref.bias = x
return np.mean(mbref.get_specific_mb(h, w, cyrs))
opti_bias = optimization.brentq(to_minimize, -10000, 10000, xtol=0.01)
mbref.bias = opti_bias
assert np.allclose(np.mean(mbref.get_specific_mb(h, w, cyrs)),
0,
atol=0.01)
rdf = pd.DataFrame(index=cyrs)
for y in cyrs:
rdf.loc[y, 'SMB'] = mbref.get_specific_mb(h, w, year=y)
t, _, p, _ = mbref.get_annual_climate([np.mean(h)], year=y)
rdf.loc[y, 'TEMP'] = t
rdf.loc[y, 'PRCP'] = p
if sigma_smb is None:
sigma_smb = rdf.std()['SMB'] / 2
if sigma_t is None:
sigma_t = rdf.std()['TEMP'] / 2
if sigma_p is None:
sigma_p = (rdf.std() / rdf.mean())['PRCP'] / 2
mb = mbmods.RandomMassBalance(gdir, all_years=True, seed=0)
mb.mbmod.temp_bias = orig_mb_tbias
mb.mbmod.prcp_bias = orig_mb_pbias
mb.mbmod.bias = orig_mb_sbias + opti_bias
rmb = mbmods.UncertainMassBalance(mb,
rdn_temp_bias_seed=rdn_temp_bias_seed,
rdn_temp_bias_sigma=sigma_t,
rdn_prcp_bias_seed=rdn_prcp_bias_seed,
rdn_prcp_bias_sigma=sigma_p,
rdn_bias_seed=rdn_bias_seed,
rdn_bias_sigma=sigma_smb)
return robust_model_run(gdir,
output_filesuffix=output_filesuffix,
mb_model=rmb,
ys=0,
ye=nyears,
**kwargs)
def run_and_store_from_disk(rgi,
histalp_storage,
commit_storage,
y0=1999,
years=300,
seed=None,
unique_samples=False,
halfsize=15):
for i in np.arange(999):
# Local working directory (where OGGM will write its output)
storage_dir = os.path.join(histalp_storage, rgi, '{:02d}'.format(i),
rgi[:8], rgi[:11], rgi)
new_dir = os.path.join(cfg.PATHS['working_dir'], 'per_glacier',
rgi[:8], rgi[:11], rgi)
# make sure directory is empty:
try:
shutil.rmtree(new_dir)
except FileNotFoundError:
pass
# if path does not exist, we handled all ensemble members:
try:
shutil.copytree(storage_dir, new_dir)
except FileNotFoundError:
log.info('processed {:02d} ensemble members'.format(i))
break
gdir = GlacierDirectory(rgi)
pdict = gdir.get_climate_info()['ensemble_calibration']
cfg.PARAMS['prcp_scaling_factor'] = pdict['prcp_scaling_factor']
default_glena = 2.4e-24
cfg.PARAMS['glen_a'] = pdict['glena_factor'] * default_glena
cfg.PARAMS['inversion_glen_a'] = pdict['glena_factor'] * default_glena
mbbias = pdict['mbbias']
tmp_mod = FileModel(
gdir.get_filepath('model_run',
filesuffix='_histalp_{:02d}'.format(i)))
tmp_mod.run_until(tmp_mod.last_yr)
mb = MultipleFlowlineMassBalance(gdir,
mb_model_class=RandomMassBalance,
filename='climate_monthly',
bias=mbbias,
y0=y0,
seed=seed,
unique_samples=unique_samples,
halfsize=halfsize)
robust_model_run(gdir,
output_filesuffix='commitment{:04d}_{:02d}'.format(
y0, i),
mb_model=mb,
ys=0,
ye=years,
init_model_fls=tmp_mod.fls)
fn1 = 'model_diagnostics_commitment{:04d}_{:02d}.nc'.format(y0, i)
shutil.copyfile(
gdir.get_filepath('model_diagnostics',
filesuffix='commitment{:04d}_{:02d}'.format(
y0, i)), os.path.join(commit_storage, fn1))
fn4 = 'model_run_commitment{:04d}_{:02d}.nc'.format(y0, i)
shutil.copyfile(
gdir.get_filepath('model_run',
filesuffix='commitment{:04d}_{:02d}'.format(
y0, i)), os.path.join(commit_storage, fn4))
def run_uncertain_random_climate(gdir, nyears=700,
output_filesuffix='',
sigma_t=None,
sigma_p=None,
sigma_smb=None,
rdn_temp_bias_seed=None,
rdn_prcp_bias_seed=None,
rdn_bias_seed=None,
orig_mb_tbias=0,
orig_mb_pbias=1,
orig_mb_sbias=0,
**kwargs):
"""Test stuff
Parameters
----------
gdir
nyears
output_filesuffix
rdn_temp_bias_seed
rdn_prcp_bias_seed
rdn_bias_seed
kwargs
Returns
-------
"""
# Find the optimal bias for a balanced MB
mbref = mbmods.PastMassBalance(gdir)
h, w = gdir.get_inversion_flowline_hw()
cyrs = np.unique(mbref.years)
def to_minimize(x):
mbref.bias = x
return np.mean(mbref.get_specific_mb(h, w, cyrs))
opti_bias = optimization.brentq(to_minimize, -10000, 10000, xtol=0.01)
mbref.bias = opti_bias
assert np.allclose(np.mean(mbref.get_specific_mb(h, w, cyrs)),
0, atol=0.01)
rdf = pd.DataFrame(index=cyrs)
for y in cyrs:
rdf.loc[y, 'SMB'] = mbref.get_specific_mb(h, w, year=y)
t, _, p, _ = mbref.get_annual_climate([np.mean(h)], year=y)
rdf.loc[y, 'TEMP'] = t
rdf.loc[y, 'PRCP'] = p
if sigma_smb is None:
sigma_smb = rdf.std()['SMB'] / 2
if sigma_t is None:
sigma_t = rdf.std()['TEMP'] / 2
if sigma_p is None:
sigma_p = (rdf.std() / rdf.mean())['PRCP'] / 2
mb = mbmods.RandomMassBalance(gdir, all_years=True, seed=0)
mb.mbmod.temp_bias = orig_mb_tbias
mb.mbmod.prcp_bias = orig_mb_pbias
mb.mbmod.bias = orig_mb_sbias + opti_bias
rmb = mbmods.UncertainMassBalance(mb,
rdn_temp_bias_seed=rdn_temp_bias_seed,
rdn_temp_bias_sigma=sigma_t,
rdn_prcp_bias_seed=rdn_prcp_bias_seed,
rdn_prcp_bias_sigma=sigma_p,
rdn_bias_seed=rdn_bias_seed,
rdn_bias_sigma=sigma_smb)
return robust_model_run(gdir, output_filesuffix=output_filesuffix,
mb_model=rmb, ys=0, ye=nyears,
**kwargs)