def test_constant_bed(self):
models = [FluxBasedModel, MUSCLSuperBeeModel]
lens = []
surface_h = []
volume = []
yrs = np.arange(1, 300, 2)
for model in models:
fls = dummy_constant_bed()
mb = LinearMassBalance(2600.)
model = model(fls, mb_model=mb, y0=0., glen_a=self.glen_a,
fs=self.fs, fixed_dt=10 * SEC_IN_DAY)
length = yrs * 0.
vol = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
assert model.yr == y
length[i] = fls[-1].length_m
vol[i] = fls[-1].volume_km3
lens.append(length)
volume.append(vol)
surface_h.append(fls[-1].surface_h.copy())
assert_allclose(lens[0][-1], lens[1][-1], atol=101)
assert_allclose(volume[0][-1], volume[1][-1], atol=3e-3)
assert rmsd(lens[0], lens[1]) < 50.
assert rmsd(volume[0], volume[1]) < 2e-3
assert rmsd(surface_h[0], surface_h[1]) < 1.0
def test_adaptive_ts(self):
models = [KarthausModel, FluxBasedModel, MUSCLSuperBeeModel]
steps = [31 * SEC_IN_DAY, None, None]
lens = []
surface_h = []
volume = []
yrs = np.arange(1, 500, 2)
for model, step in zip(models, steps):
fls = dummy_constant_bed()
mb = LinearMassBalance(2600.)
model = model(fls, mb_model=mb, glen_a=self.glen_a, fixed_dt=step)
length = yrs * 0.
vol = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
length[i] = fls[-1].length_m
vol[i] = fls[-1].volume_km3
lens.append(length)
volume.append(vol)
surface_h.append(fls[-1].surface_h.copy())
np.testing.assert_almost_equal(lens[0][-1], lens[1][-1])
np.testing.assert_allclose(volume[0][-1], volume[1][-1], atol=1e-2)
np.testing.assert_allclose(volume[0][-1], volume[2][-1], atol=1e-2)
self.assertTrue(utils.rmsd(lens[0], lens[1]) < 50.)
self.assertTrue(utils.rmsd(volume[2], volume[1]) < 1e-3)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[1]) < 5)
self.assertTrue(utils.rmsd(surface_h[1], surface_h[2]) < 5)
def test_adaptive_ts(self):
models = [
flowline.KarthausModel, flowline.FluxBasedModel,
flowline.MUSCLSuperBeeModel
]
steps = [SEC_IN_MONTH, None, None]
lens = []
surface_h = []
volume = []
yrs = np.arange(1, 500, 2)
for model, step in zip(models, steps):
fls = dummy_constant_bed()
mb = ConstantBalanceModel(2600.)
model = model(fls, mb_model=mb, glen_a=self.glen_a, fixed_dt=step)
length = yrs * 0.
vol = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
length[i] = fls[-1].length_m
vol[i] = fls[-1].volume_km3
lens.append(length)
volume.append(vol)
surface_h.append(fls[-1].surface_h.copy())
np.testing.assert_almost_equal(lens[0][-1], lens[1][-1])
np.testing.assert_allclose(volume[0][-1], volume[1][-1], atol=1e-2)
np.testing.assert_allclose(volume[0][-1], volume[2][-1], atol=1e-2)
self.assertTrue(utils.rmsd(lens[0], lens[1]) < 50.)
self.assertTrue(utils.rmsd(volume[0], volume[1]) < 1e-3)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[1]) < 5)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[2]) < 5)
def test_init_present_time_glacier(self):
gdirs = up_to_inversion()
# Inversion Results
cfg.PARAMS['invert_with_sliding'] = True
cfg.PARAMS['optimize_thick'] = True
workflow.inversion_tasks(gdirs)
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_results.csv')
df = pd.read_csv(fpath, index_col=0)
r1 = rmsd(df['ref_volume_km3'], df['oggm_volume_km3'])
r2 = rmsd(df['ref_volume_km3'], df['vas_volume_km3'])
self.assertTrue(r1 < r2)
cfg.PARAMS['invert_with_sliding'] = False
cfg.PARAMS['optimize_thick'] = False
workflow.inversion_tasks(gdirs)
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_results.csv')
df = pd.read_csv(fpath, index_col=0)
r1 = rmsd(df['ref_volume_km3'], df['oggm_volume_km3'])
r2 = rmsd(df['ref_volume_km3'], df['vas_volume_km3'])
self.assertTrue(r1 < r2)
# Init glacier
d = gdirs[0].read_pickle('inversion_params')
fs = d['fs']
glen_a = d['glen_a']
maxs = cfg.PARAMS['max_shape_param']
for gdir in gdirs:
flowline.init_present_time_glacier(gdir)
mb_mod = massbalance.ConstantMassBalanceModel(gdir)
fls = gdir.read_pickle('model_flowlines')
model = flowline.FluxBasedModel(fls, mb_model=mb_mod, y0=0.,
fs=fs, glen_a=glen_a)
_vol = model.volume_km3
_area = model.area_km2
if gdir.rgi_id in df.index:
gldf = df.loc[gdir.rgi_id]
# TODO: broken but should work
# assert_allclose(gldf['oggm_volume_km3'], _vol, rtol=0.03)
# assert_allclose(gldf['ref_area_km2'], _area, rtol=0.03)
maxo = max([fl.order for fl in model.fls])
for fl in model.fls:
self.assertTrue(np.all(fl.bed_shape > 0))
self.assertTrue(np.all(fl.bed_shape <= maxs))
if len(model.fls) > 1:
if fl.order == (maxo-1):
self.assertTrue(fl.flows_to is fls[-1])
# Test the glacier charac
dfc = utils.glacier_characteristics(gdirs)
self.assertTrue(np.all(dfc.terminus_type == 'Land-terminating'))
cc = dfc[['dem_mean_elev', 'clim_temp_avgh']].corr().values[0, 1]
self.assertTrue(cc > 0.4)
def test_set_width(self):
entity = gpd.read_file(self.rgi_file).iloc[0]
gdir = oggm.GlacierDirectory(entity, base_dir=self.testdir)
gis.define_glacier_region(gdir)
gis.glacier_masks(gdir)
centerlines.compute_centerlines(gdir)
centerlines.initialize_flowlines(gdir)
centerlines.compute_downstream_line(gdir)
centerlines.compute_downstream_bedshape(gdir)
centerlines.catchment_area(gdir)
centerlines.catchment_intersections(gdir)
centerlines.catchment_width_geom(gdir)
centerlines.catchment_width_correction(gdir)
# Test that area and area-altitude elev is fine
with utils.ncDataset(gdir.get_filepath('gridded_data')) as nc:
mask = nc.variables['glacier_mask'][:]
topo = nc.variables['topo_smoothed'][:]
rhgt = topo[np.where(mask)][:]
fls = gdir.read_pickle('inversion_flowlines')
hgt, widths = gdir.get_inversion_flowline_hw()
bs = 100
bins = np.arange(utils.nicenumber(np.min(hgt), bs, lower=True),
utils.nicenumber(np.max(hgt), bs) + 1,
bs)
h1, b = np.histogram(hgt, weights=widths, density=True, bins=bins)
h2, b = np.histogram(rhgt, density=True, bins=bins)
h1 = h1 / np.sum(h1)
h2 = h2 / np.sum(h2)
assert utils.rmsd(h1, h2) < 0.02 # less than 2% error
new_area = np.sum(widths * fls[-1].dx * gdir.grid.dx)
np.testing.assert_allclose(new_area, gdir.rgi_area_m2)
centerlines.terminus_width_correction(gdir, new_width=714)
fls = gdir.read_pickle('inversion_flowlines')
hgt, widths = gdir.get_inversion_flowline_hw()
# Check that the width is ok
np.testing.assert_allclose(fls[-1].widths[-1] * gdir.grid.dx, 714)
# Check for area distrib
bins = np.arange(utils.nicenumber(np.min(hgt), bs, lower=True),
utils.nicenumber(np.max(hgt), bs) + 1,
bs)
h1, b = np.histogram(hgt, weights=widths, density=True, bins=bins)
h2, b = np.histogram(rhgt, density=True, bins=bins)
h1 = h1 / np.sum(h1)
h2 = h2 / np.sum(h2)
assert utils.rmsd(h1, h2) < 0.02 # less than 2% error
new_area = np.sum(widths * fls[-1].dx * gdir.grid.dx)
np.testing.assert_allclose(new_area, gdir.rgi_area_m2)
def test_set_width(self):
entity = gpd.read_file(self.rgi_file).iloc[0]
gdir = oggm.GlacierDirectory(entity, base_dir=self.testdir)
gis.define_glacier_region(gdir, entity=entity)
gis.glacier_masks(gdir)
centerlines.compute_centerlines(gdir)
centerlines.initialize_flowlines(gdir)
centerlines.compute_downstream_line(gdir)
centerlines.compute_downstream_bedshape(gdir)
centerlines.catchment_area(gdir)
centerlines.catchment_intersections(gdir)
centerlines.catchment_width_geom(gdir)
centerlines.catchment_width_correction(gdir)
# Test that area and area-altitude elev is fine
with utils.ncDataset(gdir.get_filepath('gridded_data')) as nc:
mask = nc.variables['glacier_mask'][:]
topo = nc.variables['topo_smoothed'][:]
rhgt = topo[np.where(mask)][:]
fls = gdir.read_pickle('inversion_flowlines')
hgt, widths = gdir.get_inversion_flowline_hw()
bs = 100
bins = np.arange(utils.nicenumber(np.min(hgt), bs, lower=True),
utils.nicenumber(np.max(hgt), bs) + 1,
bs)
h1, b = np.histogram(hgt, weights=widths, density=True, bins=bins)
h2, b = np.histogram(rhgt, density=True, bins=bins)
h1 = h1 / np.sum(h1)
h2 = h2 / np.sum(h2)
assert utils.rmsd(h1, h2) < 0.02 # less than 2% error
new_area = np.sum(widths * fls[-1].dx * gdir.grid.dx)
np.testing.assert_allclose(new_area, gdir.rgi_area_m2)
centerlines.terminus_width_correction(gdir, new_width=714)
fls = gdir.read_pickle('inversion_flowlines')
hgt, widths = gdir.get_inversion_flowline_hw()
# Check that the width is ok
np.testing.assert_allclose(fls[-1].widths[-1] * gdir.grid.dx, 714)
# Check for area distrib
bins = np.arange(utils.nicenumber(np.min(hgt), bs, lower=True),
utils.nicenumber(np.max(hgt), bs) + 1,
bs)
h1, b = np.histogram(hgt, weights=widths, density=True, bins=bins)
h2, b = np.histogram(rhgt, density=True, bins=bins)
h1 = h1 / np.sum(h1)
h2 = h2 / np.sum(h2)
assert utils.rmsd(h1, h2) < 0.02 # less than 2% error
new_area = np.sum(widths * fls[-1].dx * gdir.grid.dx)
np.testing.assert_allclose(new_area, gdir.rgi_area_m2)
def optimize_inversion_params(gdirs):
"""Optimizes fs and fd"""
log.info('Compute the reference fs and fd parameters.')
# Get test glaciers (all glaciers with thickness data)
dfids = cfg.paths['glathida_rgi_links']
gtd_df = pd.read_csv(dfids).sort_values(by=['RGI_ID'])
dfids = gtd_df['RGI_ID'].values
ref_gdirs = [gdir for gdir in gdirs if gdir.rgi_id in dfids]
# Account for area differences between glathida and rgi
ref_area_km2 = gtd_df.RGI_AREA.values
ref_cs = gtd_df.VOLUME.values / (gtd_df.GTD_AREA.values**1.375)
ref_volume_km3 = ref_cs * ref_area_km2**1.375
ref_thickness_m = ref_volume_km3 / ref_area_km2 * 1000.
# Optimize
def to_optimize(x):
tmp_vols = np.zeros(len(ref_gdirs))
fd = 1.9e-24 * x[0]
fs = 5.7e-20 * x[1]
for i, gdir in enumerate(ref_gdirs):
v, _ = inversion_parabolic_point_slope(gdir, fs=fs, fd=fd)
tmp_vols[i] = v * 1e-9
return utils.rmsd(tmp_vols, ref_volume_km3)
out = optimization.minimize(to_optimize, [1., 1.],
bounds=((0.01, 1), (0.01, 1)),
tol=1.e-3)
# Check results and save.
fd = 1.9e-24 * out['x'][0]
fs = 5.7e-20 * out['x'][1]
tmp_vols = np.zeros(len(ref_gdirs))
for i, gdir in enumerate(ref_gdirs):
v, _ = inversion_parabolic_point_slope(gdir, fs=fs, fd=fd)
tmp_vols[i] = v * 1e-9
d = dict()
d['fs'] = fs
d['fd'] = fd
d['vol_rmsd'] = utils.rmsd(tmp_vols, ref_volume_km3)
d['thick_rmsd'] = utils.rmsd(tmp_vols / ref_area_km2 / 1000.,
ref_thickness_m)
log.info('Optimized fs={fs} and fd={fd} for a volume RMSD of '
'{vol_rmsd}'.format(**d))
df = pd.DataFrame(d, index=[0])
file = os.path.join(cfg.paths['working_dir'], 'inversion_params.csv')
df.to_csv(file)
return fs, fd
def test_init_present_time_glacier(self):
gdirs = up_to_inversion()
# Inversion Results
cfg.PARAMS['invert_with_sliding'] = True
cfg.PARAMS['optimize_thick'] = True
workflow.inversion_tasks(gdirs)
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_results.csv')
df = pd.read_csv(fpath, index_col=0)
r1 = rmsd(df['ref_volume_km3'], df['oggm_volume_km3'])
assert r1 < 0.1
cfg.PARAMS['invert_with_sliding'] = False
cfg.PARAMS['optimize_thick'] = False
workflow.inversion_tasks(gdirs)
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_results.csv')
df = pd.read_csv(fpath, index_col=0)
r1 = rmsd(df['ref_volume_km3'], df['oggm_volume_km3'])
assert r1 < 0.12
# Init glacier
d = gdirs[0].read_pickle('inversion_params')
fs = d['fs']
glen_a = d['glen_a']
for gdir in gdirs:
flowline.init_present_time_glacier(gdir)
mb_mod = massbalance.ConstantMassBalance(gdir)
fls = gdir.read_pickle('model_flowlines')
model = flowline.FluxBasedModel(fls, mb_model=mb_mod, y0=0.,
fs=fs, glen_a=glen_a)
_vol = model.volume_km3
_area = model.area_km2
if gdir.rgi_id in df.index:
gldf = df.loc[gdir.rgi_id]
assert_allclose(gldf['oggm_volume_km3'], _vol, rtol=0.05)
assert_allclose(gldf['ref_area_km2'], _area, rtol=0.05)
maxo = max([fl.order for fl in model.fls])
for fl in model.fls:
if len(model.fls) > 1:
if fl.order == (maxo-1):
self.assertTrue(fl.flows_to is fls[-1])
# Test the glacier charac
dfc = utils.glacier_characteristics(gdirs)
self.assertTrue(np.all(dfc.terminus_type == 'Land-terminating'))
cc = dfc[['flowline_mean_elev',
'tstar_avg_temp_mean_elev']].corr().values[0, 1]
assert cc < -0.8
assert np.all(dfc.t_star > 1900)
assert np.all(dfc.tstar_aar.mean() > 0.5)
def test_init_present_time_glacier(self):
gdirs = up_to_inversion()
# Inversion Results
cfg.PARAMS['invert_with_sliding'] = True
cfg.PARAMS['optimize_thick'] = True
workflow.inversion_tasks(gdirs)
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_results.csv')
df = pd.read_csv(fpath, index_col=0)
r1 = rmsd(df['ref_volume_km3'], df['oggm_volume_km3'])
r2 = rmsd(df['ref_volume_km3'], df['vas_volume_km3'])
self.assertTrue(r1 < r2)
cfg.PARAMS['invert_with_sliding'] = False
cfg.PARAMS['optimize_thick'] = False
workflow.inversion_tasks(gdirs)
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_results.csv')
df = pd.read_csv(fpath, index_col=0)
r1 = rmsd(df['ref_volume_km3'], df['oggm_volume_km3'])
r2 = rmsd(df['ref_volume_km3'], df['vas_volume_km3'])
self.assertTrue(r1 < r2)
# Init glacier
d = gdirs[0].read_pickle('inversion_params')
fs = d['fs']
glen_a = d['glen_a']
maxs = cfg.PARAMS['max_shape_param']
for gdir in gdirs:
flowline.init_present_time_glacier(gdir)
mb_mod = massbalance.ConstantMassBalanceModel(gdir)
fls = gdir.read_pickle('model_flowlines')
model = flowline.FluxBasedModel(fls,
mb_model=mb_mod,
y0=0.,
fs=fs,
glen_a=glen_a)
_vol = model.volume_km3
_area = model.area_km2
gldf = df.loc[gdir.rgi_id]
assert_allclose(gldf['oggm_volume_km3'], _vol, rtol=0.03)
assert_allclose(gldf['ref_area_km2'], _area, rtol=0.03)
maxo = max([fl.order for fl in model.fls])
for fl in model.fls:
self.assertTrue(np.all(fl.bed_shape > 0))
self.assertTrue(np.all(fl.bed_shape <= maxs))
if len(model.fls) > 1:
if fl.order == (maxo - 1):
self.assertTrue(fl.flows_to is fls[-1])
def test_tributary(self):
models = [KarthausModel, FluxBasedModel]
steps = [15 * SEC_IN_DAY, None]
flss = [dummy_width_bed(), dummy_width_bed_tributary()]
lens = []
surface_h = []
volume = []
yrs = np.arange(1, 500, 2)
for model, step, fls in zip(models, steps, flss):
mb = LinearMassBalance(2600.)
model = model(fls,
mb_model=mb,
fs=self.fs_old,
glen_a=self.aglen_old,
fixed_dt=step)
length = yrs * 0.
vol = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
assert model.yr == y
length[i] = fls[-1].length_m
vol[i] = np.sum([f.volume_km3 for f in fls])
lens.append(length)
volume.append(vol)
surface_h.append(fls[-1].surface_h.copy())
np.testing.assert_allclose(lens[0][-1], lens[1][-1], atol=101)
np.testing.assert_allclose(volume[0][-1], volume[1][-1], atol=2e-2)
np.testing.assert_allclose(utils.rmsd(lens[0], lens[1]), 0., atol=70)
np.testing.assert_allclose(utils.rmsd(volume[0], volume[1]),
0.,
atol=6e-3)
np.testing.assert_allclose(utils.rmsd(surface_h[0], surface_h[1]),
0.,
atol=5)
if do_plot: # pragma: no cover
plt.plot(lens[0], 'r')
plt.plot(lens[1], 'b')
plt.show()
plt.plot(volume[0], 'r')
plt.plot(volume[1], 'b')
plt.show()
plt.plot(fls[-1].bed_h, 'k')
plt.plot(surface_h[0], 'r')
plt.plot(surface_h[1], 'b')
plt.show()
def test_init_present_time_glacier(self):
gdirs = up_to_inversion()
# Inversion Results
cfg.PARAMS['invert_with_sliding'] = True
cfg.PARAMS['optimize_thick'] = True
workflow.inversion_tasks(gdirs)
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_results.csv')
df = pd.read_csv(fpath, index_col=0)
r1 = rmsd(df['ref_volume_km3'], df['oggm_volume_km3'])
r2 = rmsd(df['ref_volume_km3'], df['vas_volume_km3'])
self.assertTrue(r1 < r2)
cfg.PARAMS['invert_with_sliding'] = False
cfg.PARAMS['optimize_thick'] = False
workflow.inversion_tasks(gdirs)
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_results.csv')
df = pd.read_csv(fpath, index_col=0)
r1 = rmsd(df['ref_volume_km3'], df['oggm_volume_km3'])
r2 = rmsd(df['ref_volume_km3'], df['vas_volume_km3'])
self.assertTrue(r1 < r2)
# Init glacier
d = gdirs[0].read_pickle('inversion_params')
fs = d['fs']
glen_a = d['glen_a']
maxs = cfg.PARAMS['max_shape_param']
for gdir in gdirs:
flowline.init_present_time_glacier(gdir)
mb_mod = massbalance.TstarMassBalanceModel(gdir)
fls = gdir.read_pickle('model_flowlines')
model = flowline.FluxBasedModel(fls, mb_model=mb_mod, y0=0.,
fs=fs, glen_a=glen_a)
_vol = model.volume_km3
_area = model.area_km2
gldf = df.loc[gdir.rgi_id]
assert_allclose(gldf['oggm_volume_km3'], _vol, rtol=0.03)
assert_allclose(gldf['ref_area_km2'], _area, rtol=0.03)
maxo = max([fl.order for fl in model.fls])
for fl in model.fls:
self.assertTrue(np.all(fl.bed_shape > 0))
self.assertTrue(np.all(fl.bed_shape <= maxs))
if len(model.fls) > 1:
if fl.order == (maxo-1):
self.assertTrue(fl.flows_to is fls[-1])
def test_tributary(self):
models = [KarthausModel, FluxBasedModel]
steps = [15 * SEC_IN_DAY, None]
flss = [dummy_width_bed(), dummy_width_bed_tributary()]
lens = []
surface_h = []
volume = []
yrs = np.arange(1, 500, 2)
for model, step, fls in zip(models, steps, flss):
mb = LinearMassBalance(2600.)
model = model(fls, mb_model=mb, fs=self.fs_old,
glen_a=self.aglen_old,
fixed_dt=step)
length = yrs * 0.
vol = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
length[i] = fls[-1].length_m
vol[i] = np.sum([f.volume_km3 for f in fls])
lens.append(length)
volume.append(vol)
surface_h.append(fls[-1].surface_h.copy())
np.testing.assert_allclose(lens[0][-1], lens[1][-1], atol=101)
np.testing.assert_allclose(volume[0][-1], volume[1][-1], atol=2e-2)
np.testing.assert_allclose(utils.rmsd(lens[0], lens[1]), 0., atol=70)
np.testing.assert_allclose(utils.rmsd(volume[0], volume[1]), 0.,
atol=6e-3)
np.testing.assert_allclose(utils.rmsd(surface_h[0], surface_h[1]), 0.,
atol=5)
if do_plot: # pragma: no cover
plt.plot(lens[0], 'r')
plt.plot(lens[1], 'b')
plt.show()
plt.plot(volume[0], 'r')
plt.plot(volume[1], 'b')
plt.show()
plt.plot(fls[-1].bed_h, 'k')
plt.plot(surface_h[0], 'r')
plt.plot(surface_h[1], 'b')
plt.show()
def to_optimize(x):
tmp_ = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
for i, gdir in enumerate(ref_gdirs):
v, a = invert_parabolic_bed(gdir, glen_a=glen_a, fs=0.0, write=False)
tmp_[i] = v / a
return utils.rmsd(tmp_, ref_thickness_m)
def test_find_t0(self):
from oggm.tests.funcs import init_hef
from oggm.core import flowline
import pandas as pd
import matplotlib.pyplot as plt
do_plot = True
gdir = init_hef(border=80, invert_with_sliding=False)
flowline.init_present_time_glacier(gdir)
glacier = gdir.read_pickle('model_flowlines')
df = pd.read_csv(utils.get_demo_file('hef_lengths.csv'), index_col=0)
df.columns = ['Leclercq']
df = df.loc[1950:]
vol_ref = flowline.FlowlineModel(glacier).volume_km3
init_bias = 94. # so that "went too far" comes once on travis
rtol = 0.005
flowline.iterative_initial_glacier_search(gdir,
y0=df.index[0],
init_bias=init_bias,
rtol=rtol,
write_steps=True)
past_model = flowline.FileModel(gdir.get_filepath('model_run'))
vol_start = past_model.volume_km3
bef_fls = copy.deepcopy(past_model.fls)
mylen = past_model.length_m_ts()
df['oggm'] = mylen[12::12].values
df = df - df.iloc[-1]
past_model.run_until(2003)
vol_end = past_model.volume_km3
np.testing.assert_allclose(vol_ref, vol_end, rtol=0.05)
rmsd = utils.rmsd(df.Leclercq, df.oggm)
self.assertTrue(rmsd < 1000.)
if do_plot: # pragma: no cover
df.plot()
plt.ylabel('Glacier length (relative to 2003)')
plt.show()
fig = plt.figure()
lab = 'ref (vol={:.2f}km3)'.format(vol_ref)
plt.plot(glacier[-1].surface_h, 'k', label=lab)
lab = 'oggm start (vol={:.2f}km3)'.format(vol_start)
plt.plot(bef_fls[-1].surface_h, 'b', label=lab)
lab = 'oggm end (vol={:.2f}km3)'.format(vol_end)
plt.plot(past_model.fls[-1].surface_h, 'r', label=lab)
plt.plot(glacier[-1].bed_h, 'gray', linewidth=2)
plt.legend(loc='best')
plt.show()
def to_optimize(x):
tmp_vols = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
for i, gdir in enumerate(ref_gdirs):
v, _ = invert_parabolic_bed(gdir, glen_a=glen_a,
fs=0., write=False)
tmp_vols[i] = v * 1e-9
return utils.rmsd(tmp_vols, ref_volume_km3)
def to_optimize(x):
tmp_ = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
for i, gdir in enumerate(ref_gdirs):
v, a = mass_conservation_inversion(gdir, glen_a=glen_a,
fs=0., write=False)
tmp_[i] = v / a
return utils.rmsd(tmp_, ref_thickness_m)
def to_optimize(x):
tmp_vols = np.zeros(len(ref_gdirs))
fd = 1.9e-24 * x[0]
fs = 5.7e-20 * x[1]
for i, gdir in enumerate(ref_gdirs):
v, _ = inversion_parabolic_point_slope(gdir, fs=fs, fd=fd)
tmp_vols[i] = v * 1e-9
return utils.rmsd(tmp_vols, ref_volume_km3)
def to_optimize(x):
tmp_vols = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
for i, gdir in enumerate(ref_gdirs):
v, _ = invert_parabolic_bed(gdir, glen_a=glen_a,
fs=0., write=False)
tmp_vols[i] = v * 1e-9
return utils.rmsd(tmp_vols, ref_volume_km3)
def to_optimize(x):
tmp_ = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
for i, gdir in enumerate(ref_gdirs):
v, a = invert_parabolic_bed(gdir, glen_a=glen_a,
fs=0., write=False)
tmp_[i] = v / a
return utils.rmsd(tmp_, ref_thickness_m)
def test_ideal_glacier(self):
# we are making a
glen_a = cfg.A * 1
from oggm.core.models import flowline, massbalance
gdir = utils.GlacierDirectory(self.rgin, base_dir=self.testdir)
fls = self._parabolic_bed()
mbmod = massbalance.ConstantBalanceModel(2800.)
model = flowline.FluxBasedModel(fls, mb_model=mbmod, glen_a=glen_a)
model.run_until_equilibrium()
# from dummy bed
map_dx = 100.
towrite = []
for fl in model.fls:
# Distance between two points
dx = fl.dx * map_dx
# Widths
widths = fl.widths * map_dx
# Heights
hgt = fl.surface_h
# Flux
mb = mbmod.get_mb(hgt) * cfg.SEC_IN_YEAR * cfg.RHO
fl.flux = np.zeros(len(fl.surface_h))
fl.set_apparent_mb(mb)
flux = fl.flux * (map_dx**2) / cfg.SEC_IN_YEAR / cfg.RHO
pok = np.nonzero(widths > 10.)
widths = widths[pok]
hgt = hgt[pok]
flux = flux[pok]
angle = np.arctan(-np.gradient(hgt, dx)) # beware the minus sign
# Clip flux to 0
assert not np.any(flux < -0.1)
# add to output
cl_dic = dict(dx=dx,
flux=flux,
width=widths,
hgt=hgt,
slope_angle=angle,
is_last=True)
towrite.append(cl_dic)
# Write out
gdir.write_pickle(towrite, 'inversion_input', div_id=1)
v, a = inversion.invert_parabolic_bed(gdir, glen_a=glen_a)
v_km3 = v * 1e-9
a_km2 = np.sum(widths * dx) * 1e-6
v_vas = 0.034 * (a_km2**1.375)
np.testing.assert_allclose(v, model.volume_m3, rtol=0.01)
cl = gdir.read_pickle('inversion_output', div_id=1)[0]
assert utils.rmsd(cl['thick'],
model.fls[0].thick[:len(cl['thick'])]) < 10.
def to_optimize(x):
tmp_ = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
for i, gdir in enumerate(ref_gdirs):
v, a = mass_conservation_inversion(gdir,
glen_a=glen_a,
fs=0.,
write=False)
tmp_[i] = v / a
return utils.rmsd(tmp_, ref_thickness_m)
def to_optimize(x):
tmp_ref = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
for i, gdir in enumerate(ref_gdirs):
v, a = mass_conservation_inversion(gdir, glen_a=glen_a,
fs=0., write=False)
if optim_t:
tmp_ref[i] = v / a
else:
tmp_ref[i] = v * 1e-9
return utils.rmsd(tmp_ref, ref_data)
def to_optimize(x):
tmp_ref = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
for i, gdir in enumerate(ref_gdirs):
v, a = invert_parabolic_bed(gdir, glen_a=glen_a,
fs=0., write=False)
if optim_t:
tmp_ref[i] = v / a
else:
tmp_ref[i] = v * 1e-9
return utils.rmsd(tmp_ref, ref_data)
def to_optimize(x):
tmp_ref = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
for i, gdir in enumerate(ref_gdirs):
v, a = invert_parabolic_bed(gdir, glen_a=glen_a,
fs=0., write=False)
if optim_t:
tmp_ref[i] = v / a
else:
tmp_ref[i] = v * 1e-9
return utils.rmsd(tmp_ref, ref_data)
def to_optimize(x):
tmp_ref = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
for i, gdir in enumerate(ref_gdirs):
v, a = mass_conservation_inversion(gdir, glen_a=glen_a,
fs=0., write=False)
if optim_t:
tmp_ref[i] = v / a
else:
tmp_ref[i] = v * 1e-9
return utils.rmsd(tmp_ref, ref_data)
def test_find_t0(self):
gdir = init_hef(border=DOM_BORDER, invert_with_sliding=False)
flowline.init_present_time_glacier(gdir)
glacier = gdir.read_pickle('model_flowlines')
df = pd.read_csv(utils.get_demo_file('hef_lengths.csv'), index_col=0)
df.columns = ['Leclercq']
df = df.loc[1950:]
vol_ref = flowline.FlowlineModel(glacier).volume_km3
init_bias = 100. # 100 so that "went too far" comes once on travis
rtol = 0.005
flowline.find_inital_glacier(gdir,
y0=df.index[0],
init_bias=init_bias,
rtol=rtol,
write_steps=False)
past_model = gdir.read_pickle('past_model')
vol_start = past_model.volume_km3
bef_fls = copy.deepcopy(past_model.fls)
mylen = []
for y in df.index:
past_model.run_until(y)
mylen.append(past_model.fls[-1].length_m)
df['oggm'] = mylen
df = df - df.iloc[-1]
vol_end = past_model.volume_km3
np.testing.assert_allclose(vol_ref, vol_end, rtol=0.05)
rmsd = utils.rmsd(df.Leclercq, df.oggm)
self.assertTrue(rmsd < 1000.)
if do_plot: # pragma: no cover
df.plot()
plt.ylabel('Glacier length (relative to 2003)')
plt.show()
fig = plt.figure()
lab = 'ref (vol={:.2f}km3)'.format(vol_ref)
plt.plot(glacier[-1].surface_h, 'k', label=lab)
lab = 'oggm start (vol={:.2f}km3)'.format(vol_start)
plt.plot(bef_fls[-1].surface_h, 'b', label=lab)
lab = 'oggm end (vol={:.2f}km3)'.format(vol_end)
plt.plot(past_model.fls[-1].surface_h, 'r', label=lab)
plt.plot(glacier[-1].bed_h, 'gray', linewidth=2)
plt.legend(loc='best')
plt.show()
def test_ideal_glacier(self):
# we are making a
glen_a = cfg.A * 1
from oggm.core.models import flowline, massbalance
gdir = utils.GlacierDirectory(self.rgin, base_dir=self.testdir)
fls = self._parabolic_bed()
mbmod = massbalance.ConstantBalanceModel(2800.)
model = flowline.FluxBasedModel(fls, mb_model=mbmod, glen_a=glen_a)
model.run_until_equilibrium()
# from dummy bed
map_dx = 100.
towrite = []
for fl in model.fls:
# Distance between two points
dx = fl.dx * map_dx
# Widths
widths = fl.widths * map_dx
# Heights
hgt = fl.surface_h
# Flux
mb = mbmod.get_mb(hgt) * cfg.SEC_IN_YEAR * cfg.RHO
fl.flux = np.zeros(len(fl.surface_h))
fl.set_apparent_mb(mb)
flux = fl.flux * (map_dx**2) / cfg.SEC_IN_YEAR / cfg.RHO
pok = np.nonzero(widths > 10.)
widths = widths[pok]
hgt = hgt[pok]
flux = flux[pok]
angle = np.arctan(-np.gradient(hgt, dx)) # beware the minus sign
# Clip flux to 0
assert not np.any(flux < -0.1)
# add to output
cl_dic = dict(dx=dx, flux=flux, width=widths, hgt=hgt,
slope_angle=angle, is_last=True)
towrite.append(cl_dic)
# Write out
gdir.write_pickle(towrite, 'inversion_input', div_id=1)
v, a = inversion.invert_parabolic_bed(gdir, glen_a=glen_a)
v_km3 = v * 1e-9
a_km2 = np.sum(widths * dx) * 1e-6
v_vas = 0.034*(a_km2**1.375)
np.testing.assert_allclose(v, model.volume_m3, rtol=0.01)
cl = gdir.read_pickle('inversion_output', div_id=1)[0]
assert utils.rmsd(cl['thick'], model.fls[0].thick[:len(cl['thick'])]) < 10.
def test_find_t0(self):
gdir = init_hef(border=DOM_BORDER, invert_with_sliding=False)
flowline.init_present_time_glacier(gdir)
glacier = gdir.read_pickle('model_flowlines')
df = pd.read_csv(utils.get_demo_file('hef_lengths.csv'), index_col=0)
df.columns = ['Leclercq']
df = df.loc[1950:]
vol_ref = flowline.FlowlineModel(glacier).volume_km3
init_bias = 100. # 100 so that "went too far" comes once on travis
rtol = 0.005
flowline.find_inital_glacier(gdir, y0=df.index[0], init_bias=init_bias,
rtol=rtol, write_steps=False)
past_model = gdir.read_pickle('past_model')
vol_start = past_model.volume_km3
bef_fls = copy.deepcopy(past_model.fls)
mylen = []
for y in df.index:
past_model.run_until(y)
mylen.append(past_model.fls[-1].length_m)
df['oggm'] = mylen
df = df-df.iloc[-1]
vol_end = past_model.volume_km3
np.testing.assert_allclose(vol_ref, vol_end, rtol=0.05)
rmsd = utils.rmsd(df.Leclercq, df.oggm)
self.assertTrue(rmsd < 1000.)
if do_plot: # pragma: no cover
df.plot()
plt.ylabel('Glacier length (relative to 2003)')
plt.show()
fig = plt.figure()
lab = 'ref (vol={:.2f}km3)'.format(vol_ref)
plt.plot(glacier[-1].surface_h, 'k', label=lab)
lab = 'oggm start (vol={:.2f}km3)'.format(vol_start)
plt.plot(bef_fls[-1].surface_h, 'b', label=lab)
lab = 'oggm end (vol={:.2f}km3)'.format(vol_end)
plt.plot(past_model.fls[-1].surface_h, 'r', label=lab)
plt.plot(glacier[-1].bed_h, 'gray', linewidth=2)
plt.legend(loc='best')
plt.show()
def test_width(self):
hef_file = get_demo_file('Hintereisferner.shp')
rgidf = gpd.GeoDataFrame.from_file(hef_file)
# loop because for some reason indexing wont work
for index, entity in rgidf.iterrows():
gdir = oggm.GlacierDirectory(entity, base_dir=self.testdir)
gis.define_glacier_region(gdir, entity=entity)
gis.glacier_masks(gdir)
centerlines.compute_centerlines(gdir)
geometry.initialize_flowlines(gdir)
geometry.catchment_area(gdir)
geometry.catchment_width_geom(gdir)
geometry.catchment_width_correction(gdir)
area = 0.
otherarea = 0.
hgt = []
harea = []
for i in gdir.divide_ids:
cls = gdir.read_pickle('inversion_flowlines', div_id=i)
for cl in cls:
harea.extend(list(cl.widths * cl.dx))
hgt.extend(list(cl.surface_h))
area += np.sum(cl.widths * cl.dx)
nc = netCDF4.Dataset(gdir.get_filepath('gridded_data', div_id=i))
otherarea += np.sum(nc.variables['glacier_mask'][:])
nc.close()
nc = netCDF4.Dataset(gdir.get_filepath('gridded_data', div_id=0))
mask = nc.variables['glacier_mask'][:]
topo = nc.variables['topo_smoothed'][:]
nc.close()
rhgt = topo[np.where(mask)][:]
tdf = gpd.GeoDataFrame.from_file(gdir.get_filepath('outlines'))
np.testing.assert_allclose(area, otherarea, rtol=0.1)
area *= (gdir.grid.dx)**2
otherarea *= (gdir.grid.dx)**2
np.testing.assert_allclose(area * 10**-6,
np.float(tdf['AREA']),
rtol=1e-4)
# Check for area distrib
bins = np.arange(utils.nicenumber(np.min(hgt), 50, lower=True),
utils.nicenumber(np.max(hgt), 50) + 1, 50.)
h1, b = np.histogram(hgt, weights=harea, density=True, bins=bins)
h2, b = np.histogram(rhgt, density=True, bins=bins)
self.assertTrue(utils.rmsd(h1 * 100 * 50, h2 * 100 * 50) < 1)
def test_width(self):
hef_file = get_demo_file('Hintereisferner.shp')
rgidf = gpd.GeoDataFrame.from_file(hef_file)
# loop because for some reason indexing wont work
for index, entity in rgidf.iterrows():
gdir = cfg.GlacierDir(entity, base_dir=self.testdir)
gis.define_glacier_region(gdir, entity)
gis.glacier_masks(gdir)
centerlines.compute_centerlines(gdir)
geometry.initialize_flowlines(gdir)
geometry.catchment_area(gdir)
geometry.catchment_width_geom(gdir)
geometry.catchment_width_correction(gdir)
area = 0.
otherarea = 0.
hgt = []
harea = []
for i in gdir.divide_ids:
cls = gdir.read_pickle('inversion_flowlines', div_id=i)
for cl in cls:
harea.extend(list(cl.widths * cl.dx))
hgt.extend(list(cl.surface_h))
area += np.sum(cl.widths * cl.dx)
nc = netCDF4.Dataset(gdir.get_filepath('grids', div_id=i))
otherarea += np.sum(nc.variables['glacier_mask'][:])
nc.close()
nc = netCDF4.Dataset(gdir.get_filepath('grids', div_id=0))
mask = nc.variables['glacier_mask'][:]
topo = nc.variables['topo_smoothed'][:]
nc.close()
rhgt = topo[np.where(mask)][:]
tdf = gpd.GeoDataFrame.from_file(gdir.get_filepath('outlines'))
np.testing.assert_allclose(area, otherarea, rtol=0.1)
area *= (gdir.grid.dx) ** 2
otherarea *= (gdir.grid.dx) ** 2
np.testing.assert_allclose(area * 10**-6, np.float(tdf['AREA']), rtol=1e-4)
# Check for area distrib
bins = np.arange(utils.nicenumber(np.min(hgt), 50, lower=True),
utils.nicenumber(np.max(hgt), 50)+1,
50.)
h1, b = np.histogram(hgt, weights=harea, density=True, bins=bins)
h2, b = np.histogram(rhgt, density=True, bins=bins)
self.assertTrue(utils.rmsd(h1*100*50, h2*100*50) < 1)
def test_width(self):
hef_file = get_demo_file("Hintereisferner.shp")
entity = gpd.GeoDataFrame.from_file(hef_file).iloc[0]
gdir = oggm.GlacierDirectory(entity, base_dir=self.testdir)
gis.define_glacier_region(gdir, entity=entity)
gis.glacier_masks(gdir)
centerlines.compute_centerlines(gdir)
geometry.initialize_flowlines(gdir)
geometry.catchment_area(gdir)
geometry.catchment_width_geom(gdir)
geometry.catchment_width_correction(gdir)
area = 0.0
otherarea = 0.0
hgt = []
harea = []
for i in gdir.divide_ids:
cls = gdir.read_pickle("inversion_flowlines", div_id=i)
for cl in cls:
harea.extend(list(cl.widths * cl.dx))
hgt.extend(list(cl.surface_h))
area += np.sum(cl.widths * cl.dx)
with netCDF4.Dataset(gdir.get_filepath("gridded_data", div_id=i)) as nc:
otherarea += np.sum(nc.variables["glacier_mask"][:])
with netCDF4.Dataset(gdir.get_filepath("gridded_data", div_id=0)) as nc:
mask = nc.variables["glacier_mask"][:]
topo = nc.variables["topo_smoothed"][:]
rhgt = topo[np.where(mask)][:]
tdf = gpd.GeoDataFrame.from_file(gdir.get_filepath("outlines"))
np.testing.assert_allclose(area, otherarea, rtol=0.1)
area *= (gdir.grid.dx) ** 2
otherarea *= (gdir.grid.dx) ** 2
np.testing.assert_allclose(area * 10 ** -6, np.float(tdf["AREA"]), rtol=1e-4)
# Check for area distrib
bins = np.arange(utils.nicenumber(np.min(hgt), 50, lower=True), utils.nicenumber(np.max(hgt), 50) + 1, 50.0)
h1, b = np.histogram(hgt, weights=harea, density=True, bins=bins)
h2, b = np.histogram(rhgt, density=True, bins=bins)
self.assertTrue(utils.rmsd(h1 * 100 * 50, h2 * 100 * 50) < 1)
def test_noisy_bed(self):
models = [
flowline.KarthausModel, flowline.FluxBasedModel,
flowline.MUSCLSuperBeeModel
]
steps = [15 * SEC_IN_DAY, None, None]
lens = []
surface_h = []
volume = []
yrs = np.arange(1, 500, 2)
fls_orig = dummy_noisy_bed()
for model, step in zip(models, steps):
fls = copy.deepcopy(fls_orig)
mb = ConstantBalanceModel(2600.)
model = model(fls, mb_model=mb, glen_a=self.glen_a, fixed_dt=step)
length = yrs * 0.
vol = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
length[i] = fls[-1].length_m
vol[i] = fls[-1].volume_km3
lens.append(length)
volume.append(vol)
surface_h.append(fls[-1].surface_h.copy())
if do_plot: # pragma: no cover
plt.figure()
plt.plot(yrs, lens[0], 'r')
plt.plot(yrs, lens[1], 'b')
plt.plot(yrs, lens[2], 'g')
plt.title('Compare Length')
plt.xlabel('years')
plt.ylabel('[m]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(yrs, volume[0], 'r')
plt.plot(yrs, volume[1], 'b')
plt.plot(yrs, volume[2], 'g')
plt.title('Compare Volume')
plt.xlabel('years')
plt.ylabel('[km^3]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(fls[-1].bed_h, 'k')
plt.plot(surface_h[0], 'r')
plt.plot(surface_h[1], 'b')
plt.plot(surface_h[2], 'g')
plt.title('Compare Shape')
plt.xlabel('[m]')
plt.ylabel('Elevation [m]')
plt.legend(['Bed', 'Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=3)
plt.show()
np.testing.assert_allclose(lens[0][-1], lens[1][-1], atol=101)
np.testing.assert_allclose(volume[0][-1], volume[1][-1], atol=1e-2)
np.testing.assert_allclose(volume[0][-1], volume[2][-1], atol=1e-2)
self.assertTrue(utils.rmsd(lens[0], lens[1]) < 100.)
self.assertTrue(utils.rmsd(volume[0], volume[1]) < 1e-1)
self.assertTrue(utils.rmsd(volume[0], volume[2]) < 1e-1)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[1]) < 10)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[2]) < 10)
def test_varying_width(self):
"""This test is for a flowline glacier of variying width, i.e with an
accumulation area twice as wide as the tongue."""
# TODO: @alexjarosch here we should have a look at MUSCLSuperBeeModel
# set do_plot = True to see the plots
models = [
flowline.KarthausModel, flowline.FluxBasedModel,
flowline.MUSCLSuperBeeModel
]
steps = [15 * SEC_IN_DAY, None, None]
lens = []
surface_h = []
volume = []
yrs = np.arange(1, 500, 2)
for model, step in zip(models, steps):
fls = dummy_width_bed()
mb = ConstantBalanceModel(2600.)
model = model(fls, mb_model=mb, glen_a=self.glen_a, fixed_dt=step)
length = yrs * 0.
vol = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
length[i] = fls[-1].length_m
vol[i] = fls[-1].volume_km3
lens.append(length)
volume.append(vol)
surface_h.append(fls[-1].surface_h.copy())
if do_plot: # pragma: no cover
plt.figure()
plt.plot(yrs, lens[0], 'r')
plt.plot(yrs, lens[1], 'b')
plt.plot(yrs, lens[2], 'g')
plt.title('Compare Length')
plt.xlabel('years')
plt.ylabel('[m]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(yrs, volume[0], 'r')
plt.plot(yrs, volume[1], 'b')
plt.plot(yrs, volume[2], 'g')
plt.title('Compare Volume')
plt.xlabel('years')
plt.ylabel('[km^3]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(fls[-1].bed_h, 'k')
plt.plot(surface_h[0], 'r')
plt.plot(surface_h[1], 'b')
plt.plot(surface_h[2], 'g')
plt.title('Compare Shape')
plt.xlabel('[m]')
plt.ylabel('Elevation [m]')
plt.legend(['Bed', 'Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=3)
plt.show()
np.testing.assert_almost_equal(lens[0][-1], lens[1][-1])
np.testing.assert_allclose(volume[0][-1], volume[1][-1], atol=1e-2)
np.testing.assert_allclose(utils.rmsd(lens[0], lens[1]), 0., atol=50)
np.testing.assert_allclose(utils.rmsd(volume[0], volume[1]),
0.,
atol=3e-3)
np.testing.assert_allclose(utils.rmsd(surface_h[0], surface_h[1]),
0.,
atol=5)
def test_constant_bed(self):
map_dx = 100.
yrs = np.arange(1, 200, 5)
lens = []
volume = []
areas = []
surface_h = []
# Flowline case
fls = dummy_constant_bed(hmax=3000., hmin=1000., nx=200, map_dx=map_dx,
widths=1.)
mb = LinearMassBalance(2600.)
flmodel = FluxBasedModel(fls, mb_model=mb, y0=0.)
length = yrs * 0.
vol = yrs * 0.
area = yrs * 0
for i, y in enumerate(yrs):
flmodel.run_until(y)
assert flmodel.yr == y
length[i] = fls[-1].length_m
vol[i] = fls[-1].volume_km3
area[i] = fls[-1].area_km2
lens.append(length)
volume.append(vol)
areas.append(area)
surface_h.append(fls[-1].surface_h.copy())
# Make a 2D bed out of the 1D
bed_2d = np.repeat(fls[-1].bed_h, 3).reshape((fls[-1].nx, 3))
sdmodel = Upstream2D(bed_2d, dx=map_dx, mb_model=mb, y0=0.,
ice_thick_filter=None)
length = yrs * 0.
vol = yrs * 0.
area = yrs * 0
for i, y in enumerate(yrs):
sdmodel.run_until(y)
assert sdmodel.yr == y
surf_1d = sdmodel.ice_thick[:, 1]
length[i] = np.sum(surf_1d > 0) * sdmodel.dx
vol[i] = sdmodel.volume_km3 / 3
area[i] = sdmodel.area_km2 / 3
lens.append(length)
volume.append(vol)
areas.append(area)
surface_h.append(sdmodel.surface_h[:, 1])
assert_allclose(lens[0][-1], lens[1][-1], atol=101)
assert_allclose(volume[0][-1], volume[1][-1], atol=3e-3)
assert rmsd(lens[0], lens[1]) < 50.
assert rmsd(volume[0], volume[1]) < 2e-3
assert rmsd(areas[0], areas[1]) < 2e-3
assert rmsd(surface_h[0], surface_h[1]) < 1.0
# Store
run_ds = sdmodel.run_until_and_store(sdmodel.yr+50)
assert 'ice_thickness' in run_ds
# Other direction
bed_2d = np.repeat(fls[-1].bed_h, 3).reshape((fls[-1].nx, 3)).T
sdmodel = Upstream2D(bed_2d, dx=map_dx, mb_model=mb, y0=0.,
ice_thick_filter=None)
length = yrs * 0.
vol = yrs * 0.
area = yrs * 0
for i, y in enumerate(yrs):
sdmodel.run_until(y)
assert sdmodel.yr == y
surf_1d = sdmodel.ice_thick[1, :]
length[i] = np.sum(surf_1d > 0) * sdmodel.dx
vol[i] = sdmodel.volume_km3 / 3
area[i] = sdmodel.area_km2 / 3
lens.append(length)
volume.append(vol)
areas.append(area)
surface_h.append(sdmodel.surface_h[:, 1])
assert_allclose(lens[0][-1], lens[1][-1], atol=101)
assert_allclose(volume[0][-1], volume[1][-1], atol=3e-3)
assert rmsd(lens[0], lens[1]) < 50.
assert rmsd(volume[0], volume[1]) < 2e-3
assert rmsd(areas[0], areas[1]) < 2e-3
assert rmsd(surface_h[0], surface_h[1]) < 1.0
def test_first_shot(self):
df = up_to_inversion()
self.assertTrue(rmsd(df['ref_vol'], df['oggm_vol']) < rmsd(df['ref_vol'], df['vas_vol']))
def test_add_consensus(self, class_case_dir, monkeypatch):
# Init
cfg.initialize()
cfg.PARAMS['use_intersects'] = False
cfg.PATHS['working_dir'] = class_case_dir
cfg.PATHS['dem_file'] = get_demo_file('hef_srtm.tif')
entity = gpd.read_file(get_demo_file('Hintereisferner_RGI5.shp'))
entity['RGIId'] = 'RGI60-11.00897'
gdir = workflow.init_glacier_directories(entity)[0]
tasks.define_glacier_region(gdir)
tasks.glacier_masks(gdir)
ft = utils.get_demo_file('RGI60-11.00897_thickness.tif')
monkeypatch.setattr(utils, 'file_downloader', lambda x: ft)
bedtopo.add_consensus_thickness(gdir)
# Check with rasterio
cfg.add_to_basenames('consensus', 'consensus.tif')
gis.rasterio_to_gdir(gdir, ft, 'consensus', resampling='bilinear')
with xr.open_dataset(gdir.get_filepath('gridded_data')) as ds:
mine = ds.consensus_ice_thickness
with xr.open_rasterio(gdir.get_filepath('consensus')) as ds:
ref = ds.isel(band=0)
# Check area
my_area = np.sum(np.isfinite(mine.data)) * gdir.grid.dx**2
np.testing.assert_allclose(my_area, gdir.rgi_area_m2, rtol=0.07)
rio_area = np.sum(ref.data > 0) * gdir.grid.dx**2
np.testing.assert_allclose(rio_area, gdir.rgi_area_m2, rtol=0.15)
np.testing.assert_allclose(my_area, rio_area, rtol=0.15)
# They are not same:
# - interpolation not 1to1 same especially at borders
# - we preserve total volume
np.testing.assert_allclose(mine.sum(), ref.sum(), rtol=0.01)
assert utils.rmsd(ref, mine) < 2
# Check vol
cdf = pd.read_hdf(utils.get_demo_file('rgi62_itmix_df.h5'))
ref_vol = cdf.loc[gdir.rgi_id].vol_itmix_m3
my_vol = mine.sum() * gdir.grid.dx**2
np.testing.assert_allclose(my_vol, ref_vol)
# Now check the rest of the workflow
# Check that no error when var not there
vn = 'consensus_ice_thickness'
centerlines.elevation_band_flowline(gdir, bin_variables=[vn, 'foo'])
# Check vol
df = pd.read_csv(gdir.get_filepath('elevation_band_flowline'),
index_col=0)
my_vol = (df[vn] * df['area']).sum()
np.testing.assert_allclose(my_vol, ref_vol)
centerlines.fixed_dx_elevation_band_flowline(gdir,
bin_variables=[vn, 'foo'])
fdf = pd.read_csv(gdir.get_filepath('elevation_band_flowline',
filesuffix='_fixed_dx'),
index_col=0)
# Check vol
my_vol = (fdf[vn] * fdf['area_m2']).sum()
np.testing.assert_allclose(my_vol, ref_vol)
def optimize_inversion_params(gdirs):
"""Optimizes fs and fd based on GlaThiDa thicknesses.
We use the glacier averaged thicknesses provided by GlaThiDa and correct
them for differences in area with RGI, using a glacier specific volume-area
scaling formula.
Parameters
----------
gdirs: list of oggm.GlacierDirectory objects
"""
# Do we even need to do this?
if not cfg.PARAMS['optimize_inversion_params']:
log.info('User did not want to optimize the inversion params')
return
# Get test glaciers (all glaciers with thickness data)
fpath = utils.get_glathida_file()
try:
gtd_df = pd.read_csv(fpath).sort_values(by=['RGI_ID'])
except AttributeError:
gtd_df = pd.read_csv(fpath).sort(columns=['RGI_ID'])
dfids = gtd_df['RGI_ID'].values
ref_gdirs = [gdir for gdir in gdirs if gdir.rgi_id in dfids]
ref_rgiids = [gdir.rgi_id for gdir in ref_gdirs]
gtd_df = gtd_df.set_index('RGI_ID').loc[ref_rgiids]
# Account for area differences between glathida and rgi
gtd_df['RGI_AREA'] = [gdir.rgi_area_km2 for gdir in ref_gdirs]
ref_area_km2 = gtd_df.RGI_AREA.values
gtd_df.VOLUME = gtd_df.MEAN_THICKNESS * gtd_df.GTD_AREA * 1e-3
ref_cs = gtd_df.VOLUME.values / (gtd_df.GTD_AREA.values**1.375)
ref_volume_km3 = ref_cs * ref_area_km2**1.375
ref_thickness_m = ref_volume_km3 / ref_area_km2 * 1000.
# Minimize volume or thick RMSD?
optim_t = cfg.PARAMS['optimize_thick']
if optim_t:
ref_data = ref_thickness_m
tol = 0.1
else:
ref_data = ref_volume_km3
tol = 1.e-4
if cfg.PARAMS['invert_with_sliding']:
# Optimize with both params
log.info('Compute the inversion parameters.')
def to_optimize(x):
tmp_ref = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
fs = cfg.FS * x[1]
for i, gdir in enumerate(ref_gdirs):
v, a = invert_parabolic_bed(gdir, glen_a=glen_a,
fs=fs, write=False)
if optim_t:
tmp_ref[i] = v / a
else:
tmp_ref[i] = v * 1e-9
return utils.rmsd(tmp_ref, ref_data)
opti = optimization.minimize(to_optimize, [1., 1.],
bounds=((0.01, 10), (0.01, 10)),
tol=tol)
# Check results and save.
glen_a = cfg.A * opti['x'][0]
fs = cfg.FS * opti['x'][1]
else:
# Optimize without sliding
log.info('Compute the inversion parameter.')
def to_optimize(x):
tmp_ref = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
for i, gdir in enumerate(ref_gdirs):
v, a = invert_parabolic_bed(gdir, glen_a=glen_a,
fs=0., write=False)
if optim_t:
tmp_ref[i] = v / a
else:
tmp_ref[i] = v * 1e-9
return utils.rmsd(tmp_ref, ref_data)
opti = optimization.minimize(to_optimize, [1.],
bounds=((0.01, 10), ),
tol=tol)
# Check results and save.
glen_a = cfg.A * opti['x'][0]
fs = 0.
# This is for the stats
oggm_volume_m3 = np.zeros(len(ref_gdirs))
rgi_area_m2 = np.zeros(len(ref_gdirs))
for i, gdir in enumerate(ref_gdirs):
v, a = invert_parabolic_bed(gdir, glen_a=glen_a, fs=fs,
write=False)
oggm_volume_m3[i] = v
rgi_area_m2[i] = a
assert np.allclose(rgi_area_m2 * 1e-6, ref_area_km2)
# This is for each glacier
out = dict()
out['glen_a'] = glen_a
out['fs'] = fs
out['factor_glen_a'] = opti['x'][0]
try:
out['factor_fs'] = opti['x'][1]
except IndexError:
out['factor_fs'] = 0.
for gdir in gdirs:
gdir.write_pickle(out, 'inversion_params')
# This is for the working dir
# Simple stats
out['vol_rmsd'] = utils.rmsd(oggm_volume_m3 * 1e-9, ref_volume_km3)
out['thick_rmsd'] = utils.rmsd(oggm_volume_m3 * 1e-9 / ref_area_km2 / 1000,
ref_thickness_m)
log.info('Optimized glen_a and fs with a factor {factor_glen_a:.2f} and '
'{factor_fs:.2f} for a thick RMSD of '
'{thick_rmsd:.1f} and a volume RMSD of '
'{vol_rmsd:.3f}'.format(**out))
df = pd.DataFrame(out, index=[0])
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_params.csv')
df.to_csv(fpath)
# All results
df = dict()
df['ref_area_km2'] = ref_area_km2
df['ref_volume_km3'] = ref_volume_km3
df['oggm_volume_km3'] = oggm_volume_m3 * 1e-9
df['vas_volume_km3'] = 0.034*(df['ref_area_km2']**1.375)
rgi_id = [gdir.rgi_id for gdir in ref_gdirs]
df = pd.DataFrame(df, index=rgi_id)
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_results.csv')
df.to_csv(fpath)
# return value for tests
return out
print('Area % coverage where we find a k value with racmo data',
area_with_cal_result_racmo / study_area * 100)
#Uncomment for paper information
# ids_negative = df_racmo_negative.RGIId.values
# keep_ids_negative = [(i in ids_negative) for i in rgidf.RGIId]
# rgidf_racmo_negative = rgidf.iloc[keep_ids_negative]
#
# area_racmo_negative = rgidf_racmo_negative.Area.sum()
#
# print('Area % coverage where racmo data is negative',
# area_racmo_negative/ study_area * 100)
#
# negative_percet = np.round(area_racmo_negative/ study_area * 100)
RMSD = utils.rmsd(df_vel_result.u_obs, df_vel_result.u_surf)
print('RMSD between observations and oggm', RMSD)
mean_dev = utils.md(df_vel_result.u_obs, df_vel_result.u_surf)
print('mean difference between observations and oggm', mean_dev)
slope, intercept, r_value, p_value, std_err = stats.linregress(
df_vel_result.u_obs, df_vel_result.u_surf)
Num = area_with_cal_result / study_area * 100
print('N = ', Num)
print('bo = ', slope)
print(intercept)
print(r_value)
print(p_value)
print(mean_dev)
def test_varying_width(self):
"""This test is for a flowline glacier of variying width, i.e with an
accumulation area twice as wide as the tongue."""
# TODO: @alexjarosch here we should have a look at MUSCLSuperBeeModel
# set do_plot = True to see the plots
models = [KarthausModel, FluxBasedModel, MUSCLSuperBeeModel]
steps = [15 * SEC_IN_DAY, None, None]
lens = []
surface_h = []
volume = []
yrs = np.arange(1, 500, 2)
for model, step in zip(models, steps):
fls = dummy_width_bed()
mb = LinearMassBalance(2600.)
model = model(fls, mb_model=mb, glen_a=self.glen_a, fixed_dt=step)
length = yrs * 0.
vol = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
length[i] = fls[-1].length_m
vol[i] = fls[-1].volume_km3
lens.append(length)
volume.append(vol)
surface_h.append(fls[-1].surface_h.copy())
if do_plot: # pragma: no cover
plt.figure()
plt.plot(yrs, lens[0], 'r')
plt.plot(yrs, lens[1], 'b')
plt.plot(yrs, lens[2], 'g')
plt.title('Compare Length')
plt.xlabel('years')
plt.ylabel('[m]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(yrs, volume[0], 'r')
plt.plot(yrs, volume[1], 'b')
plt.plot(yrs, volume[2], 'g')
plt.title('Compare Volume')
plt.xlabel('years')
plt.ylabel('[km^3]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(fls[-1].bed_h, 'k')
plt.plot(surface_h[0], 'r')
plt.plot(surface_h[1], 'b')
plt.plot(surface_h[2], 'g')
plt.title('Compare Shape')
plt.xlabel('[m]')
plt.ylabel('Elevation [m]')
plt.legend(['Bed', 'Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=3)
plt.show()
np.testing.assert_almost_equal(lens[0][-1], lens[1][-1])
np.testing.assert_allclose(volume[0][-1], volume[1][-1], atol=2e-2)
np.testing.assert_allclose(utils.rmsd(lens[0], lens[1]), 0., atol=70)
np.testing.assert_allclose(utils.rmsd(volume[0], volume[1]), 0.,
atol=1e-2)
np.testing.assert_allclose(utils.rmsd(surface_h[0], surface_h[1]), 0.,
atol=5)
def test_min_slope(self):
""" Check what is the min slope a flowline model can produce
"""
models = [KarthausModel, FluxBasedModel, MUSCLSuperBeeModel]
kwargs = [{'fixed_dt': 3 * SEC_IN_DAY}, {}, {}]
lens = []
surface_h = []
volume = []
min_slope = []
yrs = np.arange(1, 700, 2)
for model, kw in zip(models, kwargs):
fls = dummy_constant_bed_obstacle()
mb = LinearMassBalance(2600.)
model = model(fls, mb_model=mb, y0=0., glen_a=self.glen_a, **kw)
length = yrs * 0.
vol = yrs * 0.
slope = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
fl = fls[-1]
length[i] = fl.length_m
vol[i] = fl.volume_km3
hgt = np.where(fl.thick > 0, fl.surface_h, np.NaN)
sl = np.arctan(-np.gradient(hgt, fl.dx_meter))
slope[i] = np.rad2deg(np.nanmin(sl))
lens.append(length)
volume.append(vol)
min_slope.append(slope)
surface_h.append(fls[-1].surface_h.copy())
np.testing.assert_allclose(lens[0][-1], lens[1][-1], atol=101)
np.testing.assert_allclose(volume[0][-1], volume[2][-1], atol=2e-3)
np.testing.assert_allclose(volume[1][-1], volume[2][-1], atol=5e-3)
self.assertTrue(utils.rmsd(volume[0], volume[2]) < 1e-2)
self.assertTrue(utils.rmsd(volume[1], volume[2]) < 1e-2)
if do_plot: # pragma: no cover
plt.figure()
plt.plot(yrs, lens[0], 'r')
plt.plot(yrs, lens[1], 'b')
plt.plot(yrs, lens[2], 'g')
plt.title('Compare Length')
plt.xlabel('years')
plt.ylabel('[m]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(yrs, volume[0], 'r')
plt.plot(yrs, volume[1], 'b')
plt.plot(yrs, volume[2], 'g')
plt.title('Compare Volume')
plt.xlabel('years')
plt.ylabel('[km^3]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(yrs, min_slope[0], 'r')
plt.plot(yrs, min_slope[1], 'b')
plt.plot(yrs, min_slope[2], 'g')
plt.title('Compare min slope')
plt.xlabel('years')
plt.ylabel('[degrees]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(fls[-1].bed_h, 'k')
plt.plot(surface_h[0], 'r')
plt.plot(surface_h[1], 'b')
plt.plot(surface_h[2], 'g')
plt.title('Compare Shape')
plt.xlabel('[m]')
plt.ylabel('Elevation [m]')
plt.legend(['Bed', 'Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=3)
plt.show()
def test_constant_bed(self):
map_dx = 100.
yrs = np.arange(1, 400, 5)
lens = []
volume = []
areas = []
surface_h = []
# Flowline case
fls = dummy_constant_bed(hmax=3000.,
hmin=1000.,
nx=200,
map_dx=map_dx,
widths=1.)
mb = LinearMassBalance(2600.)
flmodel = FluxBasedModel(fls, mb_model=mb, y0=0.)
length = yrs * 0.
vol = yrs * 0.
area = yrs * 0
for i, y in enumerate(yrs):
flmodel.run_until(y)
length[i] = fls[-1].length_m
vol[i] = fls[-1].volume_km3
area[i] = fls[-1].area_km2
lens.append(length)
volume.append(vol)
areas.append(area)
surface_h.append(fls[-1].surface_h.copy())
# Make a 2D bed out of the 1D
bed_2d = np.repeat(fls[-1].bed_h, 3).reshape((fls[-1].nx, 3))
sdmodel = Upstream2D(bed_2d,
dx=map_dx,
mb_model=mb,
y0=0.,
ice_thick_filter=None)
length = yrs * 0.
vol = yrs * 0.
area = yrs * 0
for i, y in enumerate(yrs):
sdmodel.run_until(y)
surf_1d = sdmodel.ice_thick[:, 1]
length[i] = np.sum(surf_1d > 0) * sdmodel.dx
vol[i] = sdmodel.volume_km3 / 3
area[i] = sdmodel.area_km2 / 3
lens.append(length)
volume.append(vol)
areas.append(area)
surface_h.append(sdmodel.surface_h[:, 1])
if do_plot:
plt.figure()
plt.plot(yrs, lens[0], 'r')
plt.plot(yrs, lens[1], 'b')
plt.title('Compare Length')
plt.xlabel('years')
plt.ylabel('[m]')
plt.legend(['Flowline', '2D'], loc=2)
plt.figure()
plt.plot(yrs, volume[0], 'r')
plt.plot(yrs, volume[1], 'b')
plt.title('Compare Volume')
plt.xlabel('years')
plt.ylabel('[km^3]')
plt.legend(['Flowline', '2D'], loc=2)
plt.figure()
plt.plot(fls[-1].bed_h, 'k')
plt.plot(surface_h[0], 'r')
plt.plot(surface_h[1], 'b')
plt.title('Compare Shape')
plt.xlabel('[m]')
plt.ylabel('Elevation [m]')
plt.legend(['Bed', 'Flowline', '2D'], loc=2)
plt.show()
np.testing.assert_almost_equal(lens[0][-1], lens[1][-1])
np.testing.assert_allclose(volume[0][-1], volume[1][-1], atol=3e-3)
self.assertTrue(utils.rmsd(lens[0], lens[1]) < 50.)
self.assertTrue(utils.rmsd(volume[0], volume[1]) < 2e-3)
self.assertTrue(utils.rmsd(areas[0], areas[1]) < 2e-3)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[1]) < 1.0)
# Equilibrium
sdmodel.run_until_equilibrium()
flmodel.run_until_equilibrium()
assert_allclose(sdmodel.volume_km3 / 3, flmodel.volume_km3, atol=2e-3)
assert_allclose(sdmodel.area_km2 / 3, flmodel.area_km2, atol=2e-3)
# Store
run_ds = sdmodel.run_until_and_store(sdmodel.yr + 50)
ts = run_ds['ice_thickness'].mean(dim=['y', 'x'])
assert_allclose(ts, ts.values[0], atol=1)
# Other direction
bed_2d = np.repeat(fls[-1].bed_h, 3).reshape((fls[-1].nx, 3)).T
sdmodel = Upstream2D(bed_2d,
dx=map_dx,
mb_model=mb,
y0=0.,
ice_thick_filter=None)
length = yrs * 0.
vol = yrs * 0.
area = yrs * 0
for i, y in enumerate(yrs):
sdmodel.run_until(y)
surf_1d = sdmodel.ice_thick[1, :]
length[i] = np.sum(surf_1d > 0) * sdmodel.dx
vol[i] = sdmodel.volume_km3 / 3
area[i] = sdmodel.area_km2 / 3
lens.append(length)
volume.append(vol)
areas.append(area)
surface_h.append(sdmodel.surface_h[:, 1])
np.testing.assert_almost_equal(lens[0][-1], lens[1][-1])
np.testing.assert_allclose(volume[0][-1], volume[1][-1], atol=3e-3)
self.assertTrue(utils.rmsd(lens[0], lens[1]) < 50.)
self.assertTrue(utils.rmsd(volume[0], volume[1]) < 2e-3)
self.assertTrue(utils.rmsd(areas[0], areas[1]) < 2e-3)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[1]) < 1.0)
# Equilibrium
sdmodel.run_until_equilibrium()
assert_allclose(sdmodel.volume_km3 / 3, flmodel.volume_km3, atol=2e-3)
assert_allclose(sdmodel.area_km2 / 3, flmodel.area_km2, atol=2e-3)
def to_optimize(x):
glen_a = cfg.A * x[0]
v, a = invert_parabolic_bed(gdir, glen_a=glen_a, fs=0.,
write=False)
return utils.rmsd(v / a, gtd_df['ref_thickness_m'].loc[gdir.rgi_id])
def optimize_inversion_params(gdirs):
"""Optimizes fs and fd based on GlaThiDa thicknesses.
We use the glacier averaged thicknesses provided by GlaThiDa and correct
them for differences in area with RGI, using a glacier specific volume-area
scaling formula.
Parameters
----------
gdirs: list of oggm.GlacierDirectory objects
"""
# Do we even need to do this?
if not cfg.PARAMS['optimize_inversion_params']:
log.info('User did not want to optimize the inversion params')
return
# Get test glaciers (all glaciers with thickness data)
fpath = utils.get_glathida_file()
try:
gtd_df = pd.read_csv(fpath).sort_values(by=['RGI_ID'])
except AttributeError:
gtd_df = pd.read_csv(fpath).sort(columns=['RGI_ID'])
dfids = gtd_df['RGI_ID'].values
ref_gdirs = [gdir for gdir in gdirs if gdir.rgi_id in dfids]
if len(ref_gdirs) == 0:
raise RuntimeError('No reference GlaThiDa glaciers. Maybe something '
'went wrong with the link list?')
ref_rgiids = [gdir.rgi_id for gdir in ref_gdirs]
gtd_df = gtd_df.set_index('RGI_ID').loc[ref_rgiids]
# Account for area differences between glathida and rgi
gtd_df['RGI_AREA'] = [gdir.rgi_area_km2 for gdir in ref_gdirs]
ref_area_km2 = gtd_df.RGI_AREA.values
ref_area_m2 = ref_area_km2 * 1e6
gtd_df.VOLUME = gtd_df.MEAN_THICKNESS * gtd_df.GTD_AREA * 1e-3
ref_cs = gtd_df.VOLUME.values / (gtd_df.GTD_AREA.values**1.375)
ref_volume_km3 = ref_cs * ref_area_km2**1.375
ref_thickness_m = ref_volume_km3 / ref_area_km2 * 1000.
# Minimize volume or thick RMSD?
optim_t = cfg.PARAMS['optimize_thick']
if optim_t:
ref_data = ref_thickness_m
tol = 0.1
else:
ref_data = ref_volume_km3
tol = 1.e-4
if cfg.PARAMS['invert_with_sliding']:
# Optimize with both params
log.info('Compute the inversion parameters.')
def to_optimize(x):
tmp_ref = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
fs = cfg.FS * x[1]
for i, gdir in enumerate(ref_gdirs):
v, a = mass_conservation_inversion(gdir, glen_a=glen_a,
fs=fs, write=False)
if optim_t:
tmp_ref[i] = v / a
else:
tmp_ref[i] = v * 1e-9
return utils.rmsd(tmp_ref, ref_data)
opti = optimization.minimize(to_optimize, [1., 1.],
bounds=((0.01, 10), (0.01, 10)),
tol=tol)
# Check results and save.
glen_a = cfg.A * opti['x'][0]
fs = cfg.FS * opti['x'][1]
else:
# Optimize without sliding
log.info('Compute the inversion parameter.')
def to_optimize(x):
tmp_ref = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
for i, gdir in enumerate(ref_gdirs):
v, a = mass_conservation_inversion(gdir, glen_a=glen_a,
fs=0., write=False)
if optim_t:
tmp_ref[i] = v / a
else:
tmp_ref[i] = v * 1e-9
return utils.rmsd(tmp_ref, ref_data)
opti = optimization.minimize(to_optimize, [1.],
bounds=((0.01, 10),),
tol=tol)
# Check results and save.
glen_a = cfg.A * opti['x'][0]
fs = 0.
# This is for the stats
oggm_volume_m3 = np.zeros(len(ref_gdirs))
rgi_area_m2 = np.zeros(len(ref_gdirs))
for i, gdir in enumerate(ref_gdirs):
v, a = mass_conservation_inversion(gdir, glen_a=glen_a, fs=fs,
write=False)
oggm_volume_m3[i] = v
rgi_area_m2[i] = a
assert np.allclose(rgi_area_m2 * 1e-6, ref_area_km2)
# This is for each glacier
out = dict()
out['glen_a'] = glen_a
out['fs'] = fs
out['factor_glen_a'] = opti['x'][0]
try:
out['factor_fs'] = opti['x'][1]
except IndexError:
out['factor_fs'] = 0.
for gdir in gdirs:
gdir.write_pickle(out, 'inversion_params')
# This is for the working dir
# Simple stats
out['vol_rmsd'] = utils.rmsd(oggm_volume_m3 * 1e-9, ref_volume_km3)
out['thick_rmsd'] = utils.rmsd(oggm_volume_m3 / ref_area_m2,
ref_thickness_m)
log.info('Optimized glen_a and fs with a factor {factor_glen_a:.2f} and '
'{factor_fs:.2f} for a thick RMSD of '
'{thick_rmsd:.1f} m and a volume RMSD of '
'{vol_rmsd:.3f} km3'.format(**out))
df = pd.DataFrame(out, index=[0])
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_params.csv')
df.to_csv(fpath)
# All results
df = dict()
df['ref_area_km2'] = ref_area_km2
df['ref_volume_km3'] = ref_volume_km3
df['oggm_volume_km3'] = oggm_volume_m3 * 1e-9
df['vas_volume_km3'] = 0.034*(df['ref_area_km2']**1.375)
rgi_id = [gdir.rgi_id for gdir in ref_gdirs]
df = pd.DataFrame(df, index=rgi_id)
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_results.csv')
df.to_csv(fpath)
# return value for tests
return out
def test_min_slope(self):
""" Check what is the min slope a flowline model can produce
"""
models = [KarthausModel, FluxBasedModel, MUSCLSuperBeeModel]
kwargs = [{'fixed_dt': 3*SEC_IN_DAY}, {}, {}]
lens = []
surface_h = []
volume = []
min_slope = []
yrs = np.arange(1, 700, 2)
for model, kw in zip(models, kwargs):
fls = dummy_constant_bed_obstacle()
mb = LinearMassBalance(2600.)
model = model(fls, mb_model=mb, y0=0., glen_a=self.glen_a,
**kw)
length = yrs * 0.
vol = yrs * 0.
slope = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
fl = fls[-1]
length[i] = fl.length_m
vol[i] = fl.volume_km3
hgt = np.where(fl.thick > 0, fl.surface_h, np.NaN)
sl = np.arctan(-np.gradient(hgt, fl.dx_meter))
slope[i] = np.rad2deg(np.nanmin(sl))
lens.append(length)
volume.append(vol)
min_slope.append(slope)
surface_h.append(fls[-1].surface_h.copy())
np.testing.assert_allclose(lens[0][-1], lens[1][-1], atol=101)
np.testing.assert_allclose(volume[0][-1], volume[2][-1], atol=2e-3)
np.testing.assert_allclose(volume[1][-1], volume[2][-1], atol=5e-3)
self.assertTrue(utils.rmsd(volume[0], volume[2]) < 1e-2)
self.assertTrue(utils.rmsd(volume[1], volume[2]) < 1e-2)
if do_plot: # pragma: no cover
plt.figure()
plt.plot(yrs, lens[0], 'r')
plt.plot(yrs, lens[1], 'b')
plt.plot(yrs, lens[2], 'g')
plt.title('Compare Length')
plt.xlabel('years')
plt.ylabel('[m]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(yrs, volume[0], 'r')
plt.plot(yrs, volume[1], 'b')
plt.plot(yrs, volume[2], 'g')
plt.title('Compare Volume')
plt.xlabel('years')
plt.ylabel('[km^3]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(yrs, min_slope[0], 'r')
plt.plot(yrs, min_slope[1], 'b')
plt.plot(yrs, min_slope[2], 'g')
plt.title('Compare min slope')
plt.xlabel('years')
plt.ylabel('[degrees]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(fls[-1].bed_h, 'k')
plt.plot(surface_h[0], 'r')
plt.plot(surface_h[1], 'b')
plt.plot(surface_h[2], 'g')
plt.title('Compare Shape')
plt.xlabel('[m]')
plt.ylabel('Elevation [m]')
plt.legend(['Bed', 'Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=3)
plt.show()
def optimize_distribute_thickness_single_glacier(gdir):
"""Tests many things
Parameters
----------
"""
# Get the ref data
i, j, lon, lat, ref_thick = get_ref_gtd_data(gdir)
grids_file = gdir.get_filepath('gridded_data')
with netCDF4.Dataset(grids_file) as nc:
glacier_mask = nc.variables['glacier_mask'][:]
pok = np.nonzero(glacier_mask[j, i])
assert len(pok[0]) > 0
i, j, ref_thick = i[pok], j[pok], ref_thick[pok]
# Make the parameter space
fac_slope = np.linspace(0, 1, 11)
fac_dis = np.linspace(0, 1, 11)
fac_c = 0.027 + np.arange(13) * 0.001
# Make the point matrix
out_mat = np.full(
(len(ref_thick), len(fac_c), len(fac_slope), len(fac_dis)), np.NaN)
out = pd.DataFrame()
t = 0
for zi, fc in enumerate(fac_c):
for yi, fs in enumerate(fac_slope):
for xi, fd in enumerate(fac_dis):
if (fs + fd) > 1:
continue
thick = distribute_thickness_vas(gdir,
reset=True,
vas_c=fc,
slope_factor=fs,
dis_factor=fd,
print_log=False)
thick = thick[j, i]
assert np.all(np.isfinite(thick))
out.loc[t, 'vas_c'] = fc
out.loc[t, 'fac_slope'] = fs
out.loc[t, 'fac_dis'] = fd
out.loc[t, 'fac_topo'] = 1 - fs - fd
out.loc[t, 'bias'] = np.mean(thick[pok] - ref_thick[pok])
out.loc[t, 'mad'] = utils.mad(ref_thick[pok], thick[pok])
out.loc[t, 'rmsd'] = utils.rmsd(ref_thick[pok], thick[pok])
out.loc[t, 'n_ref'] = len(ref_thick[pok])
out_mat[:, zi, yi, xi] = thick
t += 1
fs = os.path.join(gdir.dir, 'distribute_optim.csv')
out.to_csv(fs)
fs = os.path.join(gdir.dir, 'point_thick.nc')
dims = ['points', 'fac_c', 'fac_slope', 'fac_dis']
coords = {
'points': np.arange(len(ref_thick)),
'fac_c': fac_c,
'fac_slope': fac_slope,
'fac_dis': fac_dis
}
out_mat = xr.DataArray(out_mat, dims=dims, coords=coords)
out_mat = out_mat.to_dataset(name='thick')
out_mat['ref_thick'] = (('points', ), ref_thick)
out_mat.to_netcdf(fs)
return out, out_mat
def test_constant_bed(self):
models = [KarthausModel, FluxBasedModel, MUSCLSuperBeeModel]
lens = []
surface_h = []
volume = []
yrs = np.arange(1, 700, 2)
for model in models:
fls = dummy_constant_bed()
mb = LinearMassBalance(2600.)
model = model(fls, mb_model=mb, y0=0., glen_a=self.glen_a,
fs=self.fs, fixed_dt=10 * SEC_IN_DAY)
length = yrs * 0.
vol = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
length[i] = fls[-1].length_m
vol[i] = fls[-1].volume_km3
lens.append(length)
volume.append(vol)
surface_h.append(fls[-1].surface_h.copy())
if do_plot:
plt.figure()
plt.plot(yrs, lens[0], 'r')
plt.plot(yrs, lens[1], 'b')
plt.plot(yrs, lens[2], 'g')
plt.title('Compare Length')
plt.xlabel('years')
plt.ylabel('[m]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(yrs, volume[0], 'r')
plt.plot(yrs, volume[1], 'b')
plt.plot(yrs, volume[2], 'g')
plt.title('Compare Volume')
plt.xlabel('years')
plt.ylabel('[km^3]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(fls[-1].bed_h, 'k')
plt.plot(surface_h[0], 'r')
plt.plot(surface_h[1], 'b')
plt.plot(surface_h[2], 'g')
plt.title('Compare Shape')
plt.xlabel('[m]')
plt.ylabel('Elevation [m]')
plt.legend(['Bed', 'Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=3)
plt.show()
np.testing.assert_almost_equal(lens[0][-1], lens[1][-1])
np.testing.assert_allclose(volume[0][-1], volume[2][-1], atol=3e-3)
np.testing.assert_allclose(volume[1][-1], volume[2][-1], atol=3e-3)
self.assertTrue(utils.rmsd(lens[0], lens[2]) < 50.)
self.assertTrue(utils.rmsd(lens[1], lens[2]) < 50.)
self.assertTrue(utils.rmsd(volume[0], volume[2]) < 2e-3)
self.assertTrue(utils.rmsd(volume[1], volume[2]) < 2e-3)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[2]) < 1.0)
self.assertTrue(utils.rmsd(surface_h[1], surface_h[2]) < 1.0)
def optimize_thick(gdirs):
"""Optimizes fd based on GlaThiDa thicknesses.
We use the glacier averaged thicknesses provided by GlaThiDa and correct
them for differences in area with RGI, using a glacier specific volume-area
scaling formula.
Parameters
----------
gdirs: list of oggm.GlacierDirectory objects
"""
gtd_df = _prepare_inv(gdirs)
ref_gdirs = gtd_df['ref_gdirs']
ref_volume_km3 = gtd_df['ref_volume_km3']
ref_area_km2 = gtd_df['ref_area_km2']
ref_thickness_m = gtd_df['ref_thickness_m']
# Optimize without sliding
log.info('Compute the inversion parameter.')
def to_optimize(x):
tmp_ = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
for i, gdir in enumerate(ref_gdirs):
v, a = mass_conservation_inversion(gdir, glen_a=glen_a,
fs=0., write=False)
tmp_[i] = v / a
return utils.rmsd(tmp_, ref_thickness_m)
opti = optimization.minimize(to_optimize, [1.],
bounds=((0.01, 10), ),
tol=1.e-4)
# Check results and save.
glen_a = cfg.A * opti['x'][0]
fs = 0.
# This is for the stats
oggm_volume_m3 = np.zeros(len(ref_gdirs))
rgi_area_m2 = np.zeros(len(ref_gdirs))
for i, gdir in enumerate(ref_gdirs):
v, a = mass_conservation_inversion(gdir, glen_a=glen_a, fs=fs,
write=False)
oggm_volume_m3[i] = v
rgi_area_m2[i] = a
assert np.allclose(rgi_area_m2 * 1e-6, ref_area_km2)
# This is for each glacier
out = dict()
out['glen_a'] = glen_a
out['fs'] = fs
out['factor_glen_a'] = opti['x'][0]
try:
out['factor_fs'] = opti['x'][1]
except IndexError:
out['factor_fs'] = 0.
for gdir in gdirs:
gdir.write_pickle(out, 'inversion_params')
# This is for the working dir
# Simple stats
out['vol_rmsd'] = utils.rmsd(oggm_volume_m3 * 1e-9, ref_volume_km3)
out['thick_rmsd'] = utils.rmsd(oggm_volume_m3 / (ref_area_km2 * 1e6),
ref_thickness_m)
log.info('Optimized glen_a and fs with a factor {factor_glen_a:.2f} and '
'{factor_fs:.2f} for a thick RMSD of {thick_rmsd:.3f}'.format(
**out))
df = pd.DataFrame(out, index=[0])
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_params.csv')
df.to_csv(fpath)
# All results
df = utils.glacier_characteristics(ref_gdirs)
df['ref_area_km2'] = ref_area_km2
df['ref_volume_km3'] = ref_volume_km3
df['ref_thickness_m'] = ref_thickness_m
df['oggm_volume_km3'] = oggm_volume_m3 * 1e-9
df['oggm_thickness_m'] = oggm_volume_m3 / (ref_area_km2 * 1e6)
df['vas_volume_km3'] = 0.034*(df['ref_area_km2']**1.375)
df['vas_thickness_m'] = df['vas_volume_km3'] / ref_area_km2 * 1000
rgi_id = [gdir.rgi_id for gdir in ref_gdirs]
df = pd.DataFrame(df, index=rgi_id)
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_results.csv')
df.to_csv(fpath)
# return value for tests
return out
def test_noisy_bed(self):
models = [KarthausModel, FluxBasedModel, MUSCLSuperBeeModel]
steps = [15 * SEC_IN_DAY, None, None]
lens = []
surface_h = []
volume = []
yrs = np.arange(1, 500, 2)
fls_orig = dummy_noisy_bed()
for model, step in zip(models, steps):
fls = copy.deepcopy(fls_orig)
mb = LinearMassBalance(2600.)
model = model(fls, mb_model=mb, glen_a=self.glen_a, fixed_dt=step)
length = yrs * 0.
vol = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
length[i] = fls[-1].length_m
vol[i] = fls[-1].volume_km3
lens.append(length)
volume.append(vol)
surface_h.append(fls[-1].surface_h.copy())
if do_plot: # pragma: no cover
plt.figure()
plt.plot(yrs, lens[0], 'r')
plt.plot(yrs, lens[1], 'b')
plt.plot(yrs, lens[2], 'g')
plt.title('Compare Length')
plt.xlabel('years')
plt.ylabel('[m]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(yrs, volume[0], 'r')
plt.plot(yrs, volume[1], 'b')
plt.plot(yrs, volume[2], 'g')
plt.title('Compare Volume')
plt.xlabel('years')
plt.ylabel('[km^3]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(fls[-1].bed_h, 'k')
plt.plot(surface_h[0], 'r')
plt.plot(surface_h[1], 'b')
plt.plot(surface_h[2], 'g')
plt.title('Compare Shape')
plt.xlabel('[m]')
plt.ylabel('Elevation [m]')
plt.legend(['Bed', 'Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=3)
plt.show()
np.testing.assert_allclose(lens[0][-1], lens[1][-1], atol=101)
np.testing.assert_allclose(volume[0][-1], volume[1][-1], atol=1e-2)
np.testing.assert_allclose(volume[0][-1], volume[2][-1], atol=1e-2)
self.assertTrue(utils.rmsd(lens[0], lens[1]) < 100.)
self.assertTrue(utils.rmsd(volume[0], volume[1]) < 1e-1)
self.assertTrue(utils.rmsd(volume[0], volume[2]) < 1e-1)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[1]) < 10)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[2]) < 10)
def test_repro_to_glacier(self, class_case_dir, monkeypatch):
# Init
cfg.initialize()
cfg.PATHS['working_dir'] = class_case_dir
cfg.PARAMS['use_intersects'] = False
cfg.PATHS['dem_file'] = get_demo_file('dem_Columbia.tif')
cfg.PARAMS['border'] = 10
entity = gpd.read_file(get_demo_file('RGI60-01.10689.shp')).iloc[0]
gdir = oggm.GlacierDirectory(entity)
tasks.define_glacier_region(gdir)
tasks.glacier_masks(gdir)
# use our files
region_files = {
'ALA': {
'vx': get_demo_file('crop_ALA_G0120_0000_vx.tif'),
'vy': get_demo_file('crop_ALA_G0120_0000_vy.tif')
}
}
monkeypatch.setattr(its_live, 'region_files', region_files)
monkeypatch.setattr(utils, 'file_downloader', lambda x: x)
with warnings.catch_warnings():
warnings.filterwarnings("ignore", category=RuntimeWarning)
its_live.velocity_to_gdir(gdir)
with xr.open_dataset(gdir.get_filepath('gridded_data')) as ds:
mask = ds.glacier_mask.data.astype(bool)
vx = ds.obs_icevel_x.where(mask).data
vy = ds.obs_icevel_y.where(mask).data
vel = np.sqrt(vx**2 + vy**2)
assert np.nanmax(vel) > 2900
assert np.nanmin(vel) < 2
# We reproject with rasterio and check no big diff
cfg.BASENAMES['its_live_vx'] = ('its_live_vx.tif', '')
cfg.BASENAMES['its_live_vy'] = ('its_live_vy.tif', '')
gis.rasterio_to_gdir(gdir,
region_files['ALA']['vx'],
'its_live_vx',
resampling='bilinear')
gis.rasterio_to_gdir(gdir,
region_files['ALA']['vy'],
'its_live_vy',
resampling='bilinear')
with xr.open_rasterio(gdir.get_filepath('its_live_vx')) as da:
_vx = da.where(mask).data.squeeze()
with xr.open_rasterio(gdir.get_filepath('its_live_vy')) as da:
_vy = da.where(mask).data.squeeze()
_vel = np.sqrt(_vx**2 + _vy**2)
np.testing.assert_allclose(utils.rmsd(vel[mask], _vel[mask]),
0,
atol=40)
np.testing.assert_allclose(utils.md(vel[mask], _vel[mask]), 0, atol=8)
if DO_PLOT:
import matplotlib.pyplot as plt
smap = salem.Map(gdir.grid.center_grid, countries=False)
smap.set_shapefile(gdir.read_shapefile('outlines'))
with warnings.catch_warnings():
warnings.filterwarnings('ignore', category=RuntimeWarning)
smap.set_topography(gdir.get_filepath('dem'))
vel = np.sqrt(vx**2 + vy**2)
smap.set_data(vel)
smap.set_plot_params(cmap='Blues', vmin=None, vmax=None)
xx, yy = gdir.grid.center_grid.xy_coordinates
xx, yy = smap.grid.transform(xx, yy, crs=gdir.grid.proj)
yy = yy[2::5, 2::5]
xx = xx[2::5, 2::5]
vx = vx[2::5, 2::5]
vy = vy[2::5, 2::5]
f, ax = plt.subplots()
smap.visualize(ax=ax,
title='ITS_LIVE velocity',
cbar_title='m yr-1')
ax.quiver(xx, yy, vx, vy)
plt.show()
def test_constant_bed(self):
models = [
flowline.KarthausModel, flowline.FluxBasedModel,
flowline.MUSCLSuperBeeModel
]
lens = []
surface_h = []
volume = []
yrs = np.arange(1, 700, 2)
for model in models:
fls = dummy_constant_bed()
mb = ConstantBalanceModel(2600.)
model = model(fls,
mb_model=mb,
y0=0.,
glen_a=self.glen_a,
fs=self.fs,
fixed_dt=10 * SEC_IN_DAY)
length = yrs * 0.
vol = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
length[i] = fls[-1].length_m
vol[i] = fls[-1].volume_km3
lens.append(length)
volume.append(vol)
surface_h.append(fls[-1].surface_h.copy())
if do_plot: # pragma: no cover
plt.figure()
plt.plot(yrs, lens[0], 'r')
plt.plot(yrs, lens[1], 'b')
plt.plot(yrs, lens[2], 'g')
plt.title('Compare Length')
plt.xlabel('years')
plt.ylabel('[m]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(yrs, volume[0], 'r')
plt.plot(yrs, volume[1], 'b')
plt.plot(yrs, volume[2], 'g')
plt.title('Compare Volume')
plt.xlabel('years')
plt.ylabel('[km^3]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(fls[-1].bed_h, 'k')
plt.plot(surface_h[0], 'r')
plt.plot(surface_h[1], 'b')
plt.plot(surface_h[2], 'g')
plt.title('Compare Shape')
plt.xlabel('[m]')
plt.ylabel('Elevation [m]')
plt.legend(['Bed', 'Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=3)
plt.show()
np.testing.assert_almost_equal(lens[0][-1], lens[1][-1])
np.testing.assert_allclose(volume[0][-1], volume[2][-1], atol=3e-3)
np.testing.assert_allclose(volume[1][-1], volume[2][-1], atol=3e-3)
self.assertTrue(utils.rmsd(lens[0], lens[2]) < 50.)
self.assertTrue(utils.rmsd(lens[1], lens[2]) < 50.)
self.assertTrue(utils.rmsd(volume[0], volume[2]) < 2e-3)
self.assertTrue(utils.rmsd(volume[1], volume[2]) < 2e-3)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[2]) < 1.0)
self.assertTrue(utils.rmsd(surface_h[1], surface_h[2]) < 1.0)
def optimize_thick(gdirs):
"""Optimizes fd based on GlaThiDa thicknesses.
We use the glacier averaged thicknesses provided by GlaThiDa and correct
them for differences in area with RGI, using a glacier specific volume-area
scaling formula.
Parameters
----------
gdirs: list of oggm.GlacierDirectory objects
"""
gtd_df = _prepare_inv(gdirs)
ref_gdirs = gtd_df['ref_gdirs']
ref_volume_km3 = gtd_df['ref_volume_km3']
ref_area_km2 = gtd_df['ref_area_km2']
ref_thickness_m = gtd_df['ref_thickness_m']
# Optimize without sliding
log.info('Compute the inversion parameter.')
def to_optimize(x):
tmp_ = np.zeros(len(ref_gdirs))
glen_a = cfg.A * x[0]
for i, gdir in enumerate(ref_gdirs):
v, a = invert_parabolic_bed(gdir,
glen_a=glen_a,
fs=0.,
write=False)
tmp_[i] = v / a
return utils.rmsd(tmp_, ref_thickness_m)
opti = optimization.minimize(to_optimize, [1.],
bounds=((0.01, 10), ),
tol=1.e-4)
# Check results and save.
glen_a = cfg.A * opti['x'][0]
fs = 0.
# This is for the stats
oggm_volume_m3 = np.zeros(len(ref_gdirs))
rgi_area_m2 = np.zeros(len(ref_gdirs))
for i, gdir in enumerate(ref_gdirs):
v, a = invert_parabolic_bed(gdir, glen_a=glen_a, fs=fs, write=False)
oggm_volume_m3[i] = v
rgi_area_m2[i] = a
assert np.allclose(rgi_area_m2 * 1e-6, ref_area_km2)
# This is for each glacier
out = dict()
out['glen_a'] = glen_a
out['fs'] = fs
out['factor_glen_a'] = opti['x'][0]
try:
out['factor_fs'] = opti['x'][1]
except IndexError:
out['factor_fs'] = 0.
for gdir in gdirs:
gdir.write_pickle(out, 'inversion_params')
# This is for the working dir
# Simple stats
out['vol_rmsd'] = utils.rmsd(oggm_volume_m3 * 1e-9, ref_volume_km3)
out['thick_rmsd'] = utils.rmsd(oggm_volume_m3 / (ref_area_km2 * 1e6),
ref_thickness_m)
log.info(
'Optimized glen_a and fs with a factor {factor_glen_a:.2f} and '
'{factor_fs:.2f} for a thick RMSD of {thick_rmsd:.3f}'.format(**out))
df = pd.DataFrame(out, index=[0])
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_params.csv')
df.to_csv(fpath)
# All results
df = utils.glacier_characteristics(ref_gdirs)
df['ref_area_km2'] = ref_area_km2
df['ref_volume_km3'] = ref_volume_km3
df['ref_thickness_m'] = ref_thickness_m
df['oggm_volume_km3'] = oggm_volume_m3 * 1e-9
df['oggm_thickness_m'] = oggm_volume_m3 / (ref_area_km2 * 1e6)
df['vas_volume_km3'] = 0.034 * (df['ref_area_km2']**1.375)
df['vas_thickness_m'] = df['vas_volume_km3'] / ref_area_km2 * 1000
rgi_id = [gdir.rgi_id for gdir in ref_gdirs]
df = pd.DataFrame(df, index=rgi_id)
fpath = os.path.join(cfg.PATHS['working_dir'],
'inversion_optim_results.csv')
df.to_csv(fpath)
# return value for tests
return out
def test_constant_bed_cliff(self):
""" a test case for mass conservation in the flowline models
the idea is to introduce a cliff in the sloping bed and see
what the models do when the cliff height is changed
"""
models = [
flowline.KarthausModel, flowline.FluxBasedModel,
flowline.MUSCLSuperBeeModel
]
lens = []
surface_h = []
volume = []
yrs = np.arange(1, 700, 2)
for model in models:
fls = dummy_constant_bed_cliff()
mb = ConstantBalanceModel(2600.)
model = model(fls,
mb_model=mb,
y0=0.,
glen_a=self.glen_a,
fs=self.fs,
fixed_dt=2 * SEC_IN_DAY)
length = yrs * 0.
vol = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
length[i] = fls[-1].length_m
vol[i] = fls[-1].volume_km3
lens.append(length)
volume.append(vol)
surface_h.append(fls[-1].surface_h.copy())
if do_plot: # pragma: no cover
plt.figure()
plt.plot(yrs, lens[0], 'r')
plt.plot(yrs, lens[1], 'b')
plt.plot(yrs, lens[2], 'g')
plt.title('Compare Length')
plt.xlabel('years')
plt.ylabel('[m]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(yrs, volume[0], 'r')
plt.plot(yrs, volume[1], 'b')
plt.plot(yrs, volume[2], 'g')
plt.title('Compare Volume')
plt.xlabel('years')
plt.ylabel('[km^3]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(fls[-1].bed_h, 'k')
plt.plot(surface_h[0], 'r')
plt.plot(surface_h[1], 'b')
plt.plot(surface_h[2], 'g')
plt.title('Compare Shape')
plt.xlabel('[m]')
plt.ylabel('Elevation [m]')
plt.legend(['Bed', 'Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=3)
plt.show()
# OK, so basically, Alex's tests below show that the other models
# are wrong and produce too much mass. There is also another more
# more trivial issue with the computation of the length, I added a
# "to do" in the code.
# Unit-testing perspective:
# verify that indeed the models are wrong of more than 50%
self.assertTrue(volume[1][-1] > volume[2][-1] * 1.5)
# Karthaus is even worse
self.assertTrue(volume[0][-1] > volume[1][-1])
if False:
# TODO: this will always fail so ignore it for now
np.testing.assert_almost_equal(lens[0][-1], lens[1][-1])
np.testing.assert_allclose(volume[0][-1], volume[2][-1], atol=2e-3)
np.testing.assert_allclose(volume[1][-1], volume[2][-1], atol=2e-3)
self.assertTrue(utils.rmsd(lens[0], lens[2]) < 50.)
self.assertTrue(utils.rmsd(lens[1], lens[2]) < 50.)
self.assertTrue(utils.rmsd(volume[0], volume[2]) < 1e-3)
self.assertTrue(utils.rmsd(volume[1], volume[2]) < 1e-3)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[2]) < 1.0)
self.assertTrue(utils.rmsd(surface_h[1], surface_h[2]) < 1.0)
def test_constant_bed(self):
map_dx = 100.
yrs = np.arange(1, 400, 5)
lens = []
volume = []
areas = []
surface_h = []
# Flowline case
fls = dummy_constant_bed(hmax=3000., hmin=1000., nx=200, map_dx=map_dx,
widths=1.)
mb = LinearMassBalance(2600.)
flmodel = FluxBasedModel(fls, mb_model=mb, y0=0.)
length = yrs * 0.
vol = yrs * 0.
area = yrs * 0
for i, y in enumerate(yrs):
flmodel.run_until(y)
length[i] = fls[-1].length_m
vol[i] = fls[-1].volume_km3
area[i] = fls[-1].area_km2
lens.append(length)
volume.append(vol)
areas.append(area)
surface_h.append(fls[-1].surface_h.copy())
# Make a 2D bed out of the 1D
bed_2d = np.repeat(fls[-1].bed_h, 3).reshape((fls[-1].nx, 3))
sdmodel = Upstream2D(bed_2d, dx=map_dx, mb_model=mb, y0=0.,
ice_thick_filter=None)
length = yrs * 0.
vol = yrs * 0.
area = yrs * 0
for i, y in enumerate(yrs):
sdmodel.run_until(y)
surf_1d = sdmodel.ice_thick[:, 1]
length[i] = np.sum(surf_1d > 0) * sdmodel.dx
vol[i] = sdmodel.volume_km3 / 3
area[i] = sdmodel.area_km2 / 3
lens.append(length)
volume.append(vol)
areas.append(area)
surface_h.append(sdmodel.surface_h[:, 1])
if do_plot:
plt.figure()
plt.plot(yrs, lens[0], 'r')
plt.plot(yrs, lens[1], 'b')
plt.title('Compare Length')
plt.xlabel('years')
plt.ylabel('[m]')
plt.legend(['Flowline', '2D'], loc=2)
plt.figure()
plt.plot(yrs, volume[0], 'r')
plt.plot(yrs, volume[1], 'b')
plt.title('Compare Volume')
plt.xlabel('years')
plt.ylabel('[km^3]')
plt.legend(['Flowline', '2D'], loc=2)
plt.figure()
plt.plot(fls[-1].bed_h, 'k')
plt.plot(surface_h[0], 'r')
plt.plot(surface_h[1], 'b')
plt.title('Compare Shape')
plt.xlabel('[m]')
plt.ylabel('Elevation [m]')
plt.legend(['Bed', 'Flowline', '2D'], loc=2)
plt.show()
np.testing.assert_almost_equal(lens[0][-1], lens[1][-1])
np.testing.assert_allclose(volume[0][-1], volume[1][-1], atol=3e-3)
self.assertTrue(utils.rmsd(lens[0], lens[1]) < 50.)
self.assertTrue(utils.rmsd(volume[0], volume[1]) < 2e-3)
self.assertTrue(utils.rmsd(areas[0], areas[1]) < 2e-3)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[1]) < 1.0)
# Equilibrium
sdmodel.run_until_equilibrium()
flmodel.run_until_equilibrium()
assert_allclose(sdmodel.volume_km3 / 3, flmodel.volume_km3, atol=2e-3)
assert_allclose(sdmodel.area_km2 / 3, flmodel.area_km2, atol=2e-3)
# Store
run_ds = sdmodel.run_until_and_store(sdmodel.yr+50)
ts = run_ds['ice_thickness'].mean(dim=['y', 'x'])
assert_allclose(ts, ts.values[0], atol=1)
# Other direction
bed_2d = np.repeat(fls[-1].bed_h, 3).reshape((fls[-1].nx, 3)).T
sdmodel = Upstream2D(bed_2d, dx=map_dx, mb_model=mb, y0=0.,
ice_thick_filter=None)
length = yrs * 0.
vol = yrs * 0.
area = yrs * 0
for i, y in enumerate(yrs):
sdmodel.run_until(y)
surf_1d = sdmodel.ice_thick[1, :]
length[i] = np.sum(surf_1d > 0) * sdmodel.dx
vol[i] = sdmodel.volume_km3 / 3
area[i] = sdmodel.area_km2 / 3
lens.append(length)
volume.append(vol)
areas.append(area)
surface_h.append(sdmodel.surface_h[:, 1])
np.testing.assert_almost_equal(lens[0][-1], lens[1][-1])
np.testing.assert_allclose(volume[0][-1], volume[1][-1], atol=3e-3)
self.assertTrue(utils.rmsd(lens[0], lens[1]) < 50.)
self.assertTrue(utils.rmsd(volume[0], volume[1]) < 2e-3)
self.assertTrue(utils.rmsd(areas[0], areas[1]) < 2e-3)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[1]) < 1.0)
# Equilibrium
sdmodel.run_until_equilibrium()
assert_allclose(sdmodel.volume_km3 / 3, flmodel.volume_km3, atol=2e-3)
assert_allclose(sdmodel.area_km2 / 3, flmodel.area_km2, atol=2e-3)
def test_cliff(self):
""" a test case for mass conservation in the flowline models
the idea is to introduce a cliff in the sloping bed and see
what the models do when the cliff height is changed
"""
models = [KarthausModel, FluxBasedModel, MUSCLSuperBeeModel]
lens = []
surface_h = []
volume = []
yrs = np.arange(1, 500, 2)
for model in models:
fls = dummy_constant_bed_cliff()
mb = LinearMassBalance(2600.)
model = model(fls, mb_model=mb, y0=0., glen_a=self.glen_a,
fs=self.fs, fixed_dt=2*SEC_IN_DAY)
length = yrs * 0.
vol = yrs * 0.
for i, y in enumerate(yrs):
model.run_until(y)
length[i] = fls[-1].length_m
vol[i] = fls[-1].volume_km3
lens.append(length)
volume.append(vol)
surface_h.append(fls[-1].surface_h.copy())
if False: # pragma: no cover
plt.figure()
plt.plot(yrs, lens[0], 'r')
plt.plot(yrs, lens[1], 'b')
plt.plot(yrs, lens[2], 'g')
plt.title('Compare Length')
plt.xlabel('years')
plt.ylabel('[m]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(yrs, volume[0], 'r')
plt.plot(yrs, volume[1], 'b')
plt.plot(yrs, volume[2], 'g')
plt.title('Compare Volume')
plt.xlabel('years')
plt.ylabel('[km^3]')
plt.legend(['Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=2)
plt.figure()
plt.plot(fls[-1].bed_h, 'k')
plt.plot(surface_h[0], 'r')
plt.plot(surface_h[1], 'b')
plt.plot(surface_h[2], 'g')
plt.title('Compare Shape')
plt.xlabel('[m]')
plt.ylabel('Elevation [m]')
plt.legend(['Bed', 'Karthaus', 'Flux', 'MUSCL-SuperBee'], loc=3)
plt.show()
# OK, so basically, Alex's tests below show that the other models
# are wrong and produce too much mass. There is also another more
# more trivial issue with the computation of the length, I added a
# "to do" in the code.
# Unit-testing perspective:
# "verify" that indeed the models are wrong of more than 50%
self.assertTrue(volume[1][-1] > volume[2][-1] * 1.5)
# Karthaus is even worse
self.assertTrue(volume[0][-1] > volume[1][-1])
if False:
# TODO: this will always fail so ignore it for now
np.testing.assert_almost_equal(lens[0][-1], lens[1][-1])
np.testing.assert_allclose(volume[0][-1], volume[2][-1], atol=2e-3)
np.testing.assert_allclose(volume[1][-1], volume[2][-1], atol=2e-3)
self.assertTrue(utils.rmsd(lens[0], lens[2]) < 50.)
self.assertTrue(utils.rmsd(lens[1], lens[2]) < 50.)
self.assertTrue(utils.rmsd(volume[0], volume[2]) < 1e-3)
self.assertTrue(utils.rmsd(volume[1], volume[2]) < 1e-3)
self.assertTrue(utils.rmsd(surface_h[0], surface_h[2]) < 1.0)
self.assertTrue(utils.rmsd(surface_h[1], surface_h[2]) < 1.0)