def find_ridges(x):
if 'time' in x.dims:
x = x.isel(time=0)
return find_ridges_spherical_hessian(x,
sigma=4,
tolerance_threshold=0.0015e-3,
scheme='second_order',
return_eigvectors=True)[return_idx]
ftle_local = ftle_local.isel(time=0)
ftle = ftle.isel(time=0)
from skimage.filters import threshold_otsu, threshold_local
from skimage.segmentation import watershed
from skimage.feature import peak_local_max
from scipy import ndimage as ndi
block_size = 301
local_thresh = ftle_local.copy(data=threshold_local(ftle_local.values, block_size,
offset=-.8))
binary_local = ftle_local > local_thresh
ftle_local_high=binary_local
distance = binary_local
ridges, eigmin, eigvectors = find_ridges_spherical_hessian(ftle,
sigma=1.2,
scheme='second_order',
angle=20 * coarse_factor, return_eigvectors=True)
ridges = ridges.copy(data=skeletonize(ridges.values))
ridges.plot()
plt.show()
ridges = filter_ridges(ridges, ftle, criteria=['mean_intensity', 'major_axis_length'],
thresholds=[1.2, 30 / coarse_factor])
ridges = ridges.where(~xr.ufuncs.isnan(ridges), 0)
mslpres_coarse = mslpres_coarse.interp(latitude=ridges.latitude.values,
longitude=ridges.longitude.values,
method='linear')
dpdx = mslpres_coarse.isel(time=-1).differentiate('longitude')
dpdy = mslpres_coarse.isel(time=-1).differentiate('latitude')
days = config['lcs_time_len'] / 4
print(file)
da = xr.open_dataarray(file)
da = np.log(da) / days
ridges_l = []
for idx, time in enumerate(da.time.values):
print(f'Doing time {idx} of {da.time.values.shape[0]}')
da_ = da.sel(time=time)
# ridges, _ = find_ridges_spherical_hessian(da_,
# sigma=1,
# scheme='second_order',
# angle=20)
if relaxed:
ridges, eigmin = find_ridges_spherical_hessian(
da_,
sigma=None,
tolerance_threshold=0.5,
return_eigvectors=False)
ridges = ridges.where(eigmin < eigmin.quantile(.2)).where(
da_ > 1.5)
else:
ridges, eigmin = find_ridges_spherical_hessian(
da_,
tolerance_threshold=0.0015e-3,
sigma=None,
scheme='second_order')
ridges = ridges.where(eigmin < -.8e-11, 0)
ridges = filter_ridges(ridges,
da_,
tmp.plot()
plt.show()
def_tensor = flowmap_gradient(da, da, sigma=1.5)
def_tensor = def_tensor.stack({'points': ['latitude', 'longitude']})
def_tensor = def_tensor.dropna(dim='points')
vals = def_tensor.transpose(..., 'points').values
vals = vals.reshape([2, 2, def_tensor.shape[-1]])
def_tensor_norm_tmp = norm(vals, axis=(0, 1), ord=2)
def_tensor_norm_tmp = def_tensor.isel(derivatives=0).drop('derivatives').copy(
data=def_tensor_norm_tmp)
def_tensor_norm_tmp = def_tensor_norm_tmp.unstack('points')
def_tensor_norm_tmp.plot()
plt.show()
ridges_tmp, eivals_temp, eigvectors_tmp = find_ridges_spherical_hessian(
da, sigma=1.2, scheme='second_order', angle=50, return_eigvectors=True)
u_tmp = eigvectors_tmp.isel(eigvectors=1) * da / 30
v_tmp = eigvectors_tmp.isel(eigvectors=0) * da / 30
u_tmp_lowres = u_tmp.coarsen(latitude=5, longitude=5, boundary='trim').mean()
v_tmp_lowres = v_tmp.coarsen(latitude=5, longitude=5, boundary='trim').mean()
fig, axs = plt.subplots(
1,
2,
subplot_kw={'projection': ccrs.PlateCarree()},
)
da.plot(ax=axs[0],
transform=ccrs.PlateCarree(),
add_colorbar=False,
robust=True)
da.plot(ax=axs[1], add_colorbar=False, robust=True)
import matplotlib.pyplot as plt
from skimage.filters import hessian
import cartopy.crs as ccrs
def gen_data(xx, yy):
return np.cos(xx * np.pi/180)*np.cos(12*yy * np.pi/180)
dshape = [150, 100]
da = xr.DataArray(np.ones(dshape), dims=['latitude', 'longitude'], coords={
'latitude': np.linspace(-50, 0, 150), 'longitude': np.linspace(-90, -30, 100)}
)
xx, yy = np.meshgrid(da.longitude.values, da.latitude.values)
da = da.copy(data=gen_data(xx, yy))
ridges, eigmin, eigvectors , gradient= find_ridges_spherical_hessian(da, sigma=None,
scheme='second_order',
return_eigvectors=True)
ridges = ridges.where(eigmin<eigmin.quantile(.1))
# gradient = gradient.copy(data=np.gradient(da.values))
# gradient = gradient.unstack()
np.abs(ridges).plot()
da.plot()
plt.show()
fig, ax = plt.subplots(subplot_kw={'projection': ccrs.PlateCarree()})
da.plot.contour(ax=ax)
da.where(ridges>0).plot(add_colorbar=True, ax=ax)
ax.quiver(eigvectors.longitude.values,
eigvectors.latitude.values,
eigvectors.isel(eigvectors=0).values * eigmin.values,
transform=ccrs.PlateCarree(),
add_colorbar=False)
ax.set_title('')
plt.savefig('../tempfigs/fig3_for_scheme.png',
dpi=600,
transparent=True,
bbox_inches='tight',
pad_inches=0)
plt.close()
ftle = ftle.isel(time=0)
ridges, _ = find_ridges_spherical_hessian(ftle,
sigma=1.2,
scheme='second_order',
angle=15,
ds_wind=ds.isel(time=-1).sel(
latitude=ftle.latitude,
longitude=ftle.longitude,
method='nearest'))
valleys, _ = find_ridges_spherical_hessian(-ftle,
sigma=1.2,
scheme='second_order',
angle=15,
ds_wind=ds.isel(time=-1).sel(
latitude=ftle.latitude,
longitude=ftle.longitude,
method='nearest'))
ridges.plot()
plt.show()
# ridges = ridges.where(ridges < -9e-10, 0) # Warning: sensitive to sigma
# ridges = ridges.where(ridges >= -9e-10, 1)
import cartopy.crs as ccrs
import matplotlib.pyplot as plt
import cmasher as cmr
import numpy as np
datapath = '/home/gab/phd/data/case_study_helen/'
files = glob.glob(datapath + 'ftle_1989*')
da = xr.open_mfdataset(files).to_array().isel(variable=0).drop('variable').load()
ridges_list = []
for idx, time in enumerate(da.time.values):
da_ = da.isel(time=idx)
da_ = np.log(da_)
ridges, eigmin, ridges_, eigvectors, gradient, angle = find_ridges_spherical_hessian(da_, sigma=None,
tolerance_threshold=0.5,
return_eigvectors=True)
ridges_valid = ridges.where(eigmin < eigmin.quantile(.2)).where(da_ > 1.5)
fig, ax = plt.subplots(1, 1, subplot_kw={'projection': ccrs.NearsidePerspective(0, 45)})
da_.plot(transform=ccrs.PlateCarree(), ax=ax, cmap=cmr.freeze, vmin=0, vmax=3, add_colorbar=False)
ridges_valid.where(xr.ufuncs.isnan(ridges_valid), 1).plot(transform=ccrs.PlateCarree(), alpha=.5,
ax=ax, cmap='Reds', add_colorbar=False)
ax.coastlines(color='red')
plt.savefig(f'figs/{idx:02d}.png', dpi=400)
plt.close()
ridges_list.append(ridges_valid)
ridges = xr.concat(ridges_list, dim='time').drop('quantile')
ridges = ridges.where(xr.ufuncs.isnan(ridges), 1)
ridges.isel(time=0).plot()
preprocess=lambda x: x.isel(time=[-1])).assign_coords(time=das.time.values)
y_traj = xr.open_mfdataset(
y_traj_files,
preprocess=lambda x: x.isel(time=[-1])).assign_coords(time=das.time.values)
das = das['__xarray_dataarray_variable__'].load()
for time in das.time.values:
ftle_ = .5 * np.log(das.sel(time=time))
(ftle_).plot(
vmin=0,
vmax=3,
)
plt.show()
x_traj_ = x_traj.sel(time=time)['u']
y_traj_ = y_traj.sel(time=time)['u']
protoridges, eigmin = find_ridges_spherical_hessian(
ftle_, tolerance_threshold=0.01e-3, sigma=None, scheme='second_order')
ridges = protoridges.where(eigmin < -1.2e-10, 0)
ridges = ridges.where(ftle_ > 1.2, 0)
ridges = ridges.where(ridges == 0, 1)
ridges = filter_ridges(ridges,
ftle_,
criteria=[
'mean_intensity',
'area',
],
thresholds=[.8, 20])
# ridges = ridges.where(ftle_ > .9)
# ridges.where(ftle_>1).plot()
# plt.show()