def get_sorted_krg_wts(vg_str, rnd_pts, abs_thresh_wt):
ord_krg_cls = OrdinaryKriging(xi=rnd_pts[:, 0],
yi=rnd_pts[:, 1],
zi=np.zeros(rnd_pts.shape[0]),
xk=np.array([0.0]),
yk=np.array([0.0]),
model=vg_str)
ord_krg_cls.krige()
wts = ord_krg_cls.lambdas.ravel()
wts.sort()
if abs_thresh_wt:
abs_wts = np.abs(wts)
abs_wts_sum = abs_wts.sum()
rel_wts = abs_wts / abs_wts_sum
zero_idxs = rel_wts <= abs_thresh_wt
non_zero_wts = wts[~zero_idxs]
sclr = 1 / non_zero_wts.sum()
assert zero_idxs.sum() < zero_idxs.size
wts[zero_idxs] = 0.0
wts[~zero_idxs] *= sclr
assert (wts[1:] - wts[:-1]).min() >= 0.0
assert np.isclose(wts.sum(), 1.0)
return wts
xlabel = 'Netatmo observed values'
ylabel = 'Netatmo interpolated values using DWD data'
measured_stns = 'Netatmo'
used_stns = 'DWD'
plot_title_acc = '_using_DWD_stations_to_find_Netatmo_values_'
ordinary_kriging = OrdinaryKriging(
xi=x_dwd,
yi=y_dwd,
zi=dwd_vals,
xk=x_netatmo,
yk=y_netatmo,
model=vgs_model)
try:
ordinary_kriging.krige()
except Exception as msg:
print('Error while Kriging', msg)
continue
# print('\nDistances are:\n', ordinary_kriging.in_dists)
# print('\nVariances are:\n', ordinary_kriging.in_vars)
# print('\nRight hand sides are:\n', ordinary_kriging.rhss)
# print('\nzks are:', ordinary_kriging.zk)
# print('\nest_vars are:\n', ordinary_kriging.est_vars)
# print('\nlambdas are:\n', ordinary_kriging.lambdas)
# print('\nmus are:\n', ordinary_kriging.mus)
# print('\n\n')
# interpolated vals
interpolated_vals = ordinary_kriging.zk
yi=dwd_ycoords,
zi=edf_dwd_vals,
xk=x_dwd_interpolate,
yk=y_dwd_interpolate,
model=vgs_model_dwd)
ordinary_kriging_netatmo_only = OrdinaryKriging(
xi=netatmo_xcoords,
yi=netatmo_ycoords,
zi=edf_netatmo_vals,
xk=x_dwd_interpolate,
yk=y_dwd_interpolate,
model=vgs_model_dwd)
try:
ordinary_kriging_dwd_netatmo_comb.krige()
ordinary_kriging_dwd_only.krige()
ordinary_kriging_netatmo_only.krige()
except Exception as msg:
print('Error while Kriging', msg)
interpolated_vals_dwd_netatmo = ordinary_kriging_dwd_netatmo_comb.zk.copy(
)
interpolated_vals_dwd_only = ordinary_kriging_dwd_only.zk.copy(
)
interpolated_vals_netatmo_only = ordinary_kriging_netatmo_only.zk.copy(
)
if interpolated_vals_dwd_netatmo < 0:
interpolated_vals_dwd_netatmo = np.nan
netatmo_dwd_y_coords = np.concatenate(
[netatmo_ycoords, dwd_ycoords])
netatmo_dwd_ppt_vals_gd = np.round(
np.hstack((ppt_netatmo_vals_gd, ppt_dwd_vals)), 2).ravel()
#print('\n+*-KRIGING WITH 1st and 2nd Filter-*+')
ordinary_kriging_dwd_netatmo_ppt = OrdinaryKriging(
xi=netatmo_dwd_x_coords,
yi=netatmo_dwd_y_coords,
zi=netatmo_dwd_ppt_vals_gd,
xk=x_dwd_interpolate,
yk=y_dwd_interpolate,
model=vgs_model_dwd_ppt)
try:
ordinary_kriging_dwd_netatmo_ppt.krige()
except Exception as msg:
print('ERROR BEI OK DWD-NETATMO', msg)
interpolated_vals_dwd_netatmo = (
ordinary_kriging_dwd_netatmo_ppt.zk.copy())
# put negative values to 0
interpolated_vals_dwd_netatmo[
interpolated_vals_dwd_netatmo < 0] = 0
#======================================================
# # SAVING PPT
#======================================================
df_interpolated_dwd_netatmos_comb.loc[
event_date, stn_comb] = np.round(interpolated_vals_dwd_netatmo,
#===============================================================
# # apply on event filter
#===============================================================
#print('NETATMO 2nd Filter')
ordinary_kriging_filter_netamto = OrdinaryKriging(
xi=dwd_xcoords,
yi=dwd_ycoords,
zi=edf_dwd_vals,
xk=netatmo_xcoords,
yk=netatmo_ycoords,
model=vgs_model_dwd_ppt)
try:
ordinary_kriging_filter_netamto.krige()
except Exception as msg:
print('Error while Error Kriging', msg)
# interpolated vals
interpolated_vals = ordinary_kriging_filter_netamto.zk
# calcualte standard deviation of estimated values
std_est_vals = np.sqrt(ordinary_kriging_filter_netamto.est_vars)
# calculate difference observed and estimated # values
try:
diff_obsv_interp = np.abs(netatmo_edf.values - interpolated_vals)
except Exception:
print('ERROR 2nd FILTER')
if ('Nug' in vgs_model_dwd or len(vgs_model_dwd) > 0) and (
'Exp' in vgs_model_dwd or 'Sph' in vgs_model_dwd):
print('**Changed Variogram model to**\n', vgs_model_dwd)
print('\n+++ KRIGING +++\n')
ordinary_kriging_dwd_only = OrdinaryKriging(
xi=dwd_xcoords,
yi=dwd_ycoords,
zi=ppt_dwd_vals,
xk=x_dwd_interpolate,
yk=y_dwd_interpolate,
model=vgs_model_dwd)
try:
ordinary_kriging_dwd_only.krige()
except Exception as msg:
print('Error while Kriging', msg)
interpolated_vals_dwd_only = ordinary_kriging_dwd_only.zk.copy(
)
print('**Interpolated DWD: ', interpolated_vals_dwd_only)
if interpolated_vals_dwd_only < 0:
interpolated_vals_dwd_only = np.nan
else:
print('no good variogram found, adding nans to df')
interpolated_vals_dwd_only = np.nan
if use_temporal_filter_after_kriging:
print('using DWD stations to find Netatmo values')
print('apllying on event filter')
ordinary_kriging_filter_netamto = OrdinaryKriging(
xi=dwd_xcoords,
yi=dwd_ycoords,
zi=edf_dwd_vals,
xk=netatmo_xcoords,
yk=netatmo_ycoords,
model=vgs_model_dwd)
try:
ordinary_kriging_filter_netamto.krige()
except Exception as msg:
print('Error while Error Kriging', msg)
continue
# interpolated vals
interpolated_vals = ordinary_kriging_filter_netamto.zk
# calcualte standard deviation of estimated values
std_est_vals = np.sqrt(
ordinary_kriging_filter_netamto.est_vars)
# calculate difference observed and estimated values
diff_obsv_interp = np.abs(
edf_netatmo_vals - interpolated_vals)
#======================================================
def krige(self, n_krige_intervals):
assert self._epsg_set_flag
assert self._bck_shp_set_flag
assert self._coords_arr_set_flag
assert self._sel_cps_arr_set_flag
assert self._cp_rules_arr_set_flag
assert self._anom_arr_set_flag
assert self._other_prms_set_flag
assert isinstance(n_krige_intervals, int)
assert n_krige_intervals > 0
assert self.sel_cps_arr.shape[0] == self.anom_arr.shape[0]
assert self.anom_arr.shape[1] == self.n_pts
self.n_krige_intervals = n_krige_intervals
self._prep_coords()
self.best_cps_mean_anoms = np.empty(shape=(self.n_cps, self.n_pts),
dtype=float)
self.best_cps_std_anoms = np.empty(shape=(self.n_cps,
self.y_coords.shape[0],
self.x_coords.shape[0]),
dtype=float)
self.best_cps_min_anoms = self.best_cps_std_anoms.copy()
self.best_cps_max_anoms = self.best_cps_std_anoms.copy()
(self.x_coords_mesh,
self.y_coords_mesh) = np.meshgrid(self.x_coords,
self.y_coords)
(self.x_coords_mesh_rav,
self.y_coords_mesh_rav) = (self.x_coords_mesh.ravel(),
self.y_coords_mesh.ravel())
assert self.x_coords_mesh_rav.shape[0] == self.n_pts
for j in range(self.n_cps):
_ = self.sel_cps_arr == j
_ = self.anom_arr[_]
_1 = _.mean(axis=0)
_2 = _.std(axis=0)
_3 = _.min(axis=0)
_4 = _.max(axis=0)
self.best_cps_mean_anoms[j] = _1
self.best_cps_std_anoms[j] = _2.reshape((self.y_coords.shape[0],
self.x_coords.shape[0]))
self.best_cps_min_anoms[j] = _3.reshape((self.y_coords.shape[0],
self.x_coords.shape[0]))
self.best_cps_max_anoms[j] = _4.reshape((self.y_coords.shape[0],
self.x_coords.shape[0]))
self.krige_z_coords = np.zeros((self.n_cps, self.n_pts_krige))
self.krige_z_coords_mesh = np.zeros((self.n_cps,
*self.krige_pts_shape))
self.cp_x_coords_list = []
self.cp_y_coords_list = []
self.cp_z_coords_list = []
self.vgs_list = []
for j in range(self.n_cps):
if self.msgs:
print('Kriging CP:', (j))
curr_mean_cp = self.best_cps_mean_anoms[j]
if self.anom_type != 'd':
curr_false_idxs = self.cp_rules_arr[j] == self.n_fuzz_nos
else:
curr_false_idxs = (
np.zeros(self.best_cps_mean_anoms[j].shape[0], dtype=bool))
curr_true_idxs = np.logical_not(curr_false_idxs)
curr_mean_cp[curr_false_idxs] = np.nan
curr_cp_vals = curr_mean_cp[curr_true_idxs]
curr_x_coord_vals = self.x_coords_mesh_rav[curr_true_idxs]
curr_y_coord_vals = self.y_coords_mesh_rav[curr_true_idxs]
curr_mean_cp = curr_mean_cp.reshape(self.x_coords_mesh.shape)
#==================================================================
# Reproject to out_epsg
#==================================================================
curr_re_x_coords, curr_re_y_coords = change_pts_crs(
curr_x_coord_vals,
curr_y_coord_vals,
self.anom_epsg,
self.out_epsg)
self.cp_x_coords_list.append(curr_re_x_coords)
self.cp_y_coords_list.append(curr_re_y_coords)
self.cp_z_coords_list.append(curr_cp_vals)
#==================================================================
# Fit Variogram
#==================================================================
variogram = Variogram(
x=curr_re_x_coords,
y=curr_re_y_coords,
z=curr_cp_vals,
perm_r_list=[1],
fit_vgs=['Sph'])
variogram.fit()
fit_vg = variogram.vg_str_list[0]
assert fit_vg
self.vgs_list.append(fit_vg)
if self.msgs:
print('Fitted variogram is:', fit_vg)
#==================================================================
# Krige
#==================================================================
ord_krig = OrdinaryKriging(xi=curr_re_x_coords,
yi=curr_re_y_coords,
zi=curr_cp_vals,
xk=self.krige_x_coords,
yk=self.krige_y_coords,
model=fit_vg)
ord_krig.krige()
krige_z_coords = ord_krig.zk
self.krige_z_coords[j] = krige_z_coords
assert check_nans_finite(krige_z_coords)
krige_z_coords_mesh = krige_z_coords.reshape(self.krige_pts_shape)
self.krige_z_coords_mesh[j] = krige_z_coords_mesh
self._kriged_flag = True
return
