Guide and example on how to use nested grids with DataTrees#
This is an example on how to use PyDDA’s ability to handle nested grids using xarray DataTrees. In this example, we load radars with two pre-generated Cf/Radial grid. The fine grids are higher resolution grids that are contained within the coarser grid.
The DataTree structure that PyDDA follows is:
::
root
|---nest_0/radar_1
|---nest_0/radar_2
|---nest_0/radar_n
|---nest_1/radar_1
|---nest_1/radar_2
|---nest_1/radar_m
Each member of this tree is a DataTree itself. PyDDA will know if the
DataTree contains data from a radar when the name of the node begins
with radar_. The root node of each grid level, in this example,
root and inner_nest will contain the keyword arguments that are
inputs to :code:pydda.retrieval.get_dd_wind_field as attributes for the
tree. PyDDA will use the attributes at each level as the arguments for the
retrieval, allowing the user to vary the coefficients by grid level.
Using :code:pydda.retrieval.get_dd_wind_field_nested will allow PyDDA
to perform the retrieval on the 0th grid first. It will then
perform on the subsequent grid levels, using the previous nest as both the
horizontal boundary conditions and initialization for the retrieval in the next
nest. Finally, PyDDA will update the winds in the first grid by nearest-
neighbor interpolation of the latter grid into the overlapping portion between
the inner and outer grid level.
PyDDA will then return the retrieved wind fields as the “u”, “v”, and “w” DataArrays inside each of the root nodes for each level, in this case root and inner_nest.
## Do imports
import pydda
import matplotlib.pyplot as plt
import warnings
from xarray import DataTree
warnings.filterwarnings("ignore")
"""
We will load pregenerated grids for this case.
"""
test_coarse0 = pydda.io.read_grid(pydda.tests.get_sample_file("test_coarse0.nc"))
test_coarse1 = pydda.io.read_grid(pydda.tests.get_sample_file("test_coarse1.nc"))
test_fine0 = pydda.io.read_grid(pydda.tests.get_sample_file("test_fine0.nc"))
test_fine1 = pydda.io.read_grid(pydda.tests.get_sample_file("test_fine1.nc"))
"""
Initalize with a zero wind field. We have HRRR data already generated for this case inside
the example data files to provide a model constraint.
"""
test_coarse0 = pydda.initialization.make_constant_wind_field(
test_coarse0, (0.0, 0.0, 0.0)
)
"""
Specify the retrieval parameters at each level
"""
kwargs_dict = dict(
Cm=256.0,
Co=1e-2,
Cx=50.0,
Cy=50.0,
Cz=50.0,
Cmod=1e-5,
model_fields=["hrrr"],
refl_field="DBZ",
wind_tol=0.5,
max_iterations=150,
engine="scipy",
)
"""
Enforce equal times for each grid. This is required for the DataTree structure since time is an
inherited dimension.
"""
test_coarse1["time"] = test_coarse0["time"]
test_fine0["time"] = test_coarse0["time"]
test_fine1["time"] = test_coarse1["time"]
"""
Provide the overlying grid structure as specified above.
"""
tree_dict = {
"/nest_0/radar_ktlx": test_coarse0,
"/nest_0/radar_kict": test_coarse1,
"/nest_1/radar_ktlx": test_fine0,
"/nest_1/radar_kict": test_fine1,
}
tree = DataTree.from_dict(tree_dict)
tree["/nest_0/"].attrs = kwargs_dict
tree["/nest_1/"].attrs = kwargs_dict
"""
Perform the retrieval
"""
grid_tree = pydda.retrieval.get_dd_wind_field_nested(tree)
"""
Plot the coarse grid output and finer grid output
"""
fig, ax = plt.subplots(1, 2, figsize=(10, 5))
pydda.vis.plot_horiz_xsection_quiver(
grid_tree["nest_0"],
ax=ax[0],
level=5,
cmap="ChaseSpectral",
vmin=-10,
vmax=80,
quiverkey_len=10.0,
background_field="DBZ",
bg_grid_no=1,
w_vel_contours=[1, 2, 5, 10],
quiver_spacing_x_km=50.0,
quiver_spacing_y_km=50.0,
quiverkey_loc="bottom_right",
)
pydda.vis.plot_horiz_xsection_quiver(
grid_tree["nest_1"],
ax=ax[1],
level=5,
cmap="ChaseSpectral",
vmin=-10,
vmax=80,
quiverkey_len=10.0,
background_field="DBZ",
bg_grid_no=1,
w_vel_contours=[1, 2, 5, 10],
quiver_spacing_x_km=50.0,
quiver_spacing_y_km=50.0,
quiverkey_loc="bottom_right",
)
plt.show()
## You are using the Python ARM Radar Toolkit (Py-ART), an open source
## library for working with weather radar data. Py-ART is partly supported
## by the U.S. Department of Energy Office of Science as part of
## the Atmospheric Radiation Measurement (ARM) User Facility.
##
## If you use this software to prepare a publication, please cite:
##
## JJ Helmus and SM Collis, JORS 2016, doi: 10.5334/jors.119
Welcome to PyDDA 2.5.0
If you are using PyDDA in your publications, please cite:
Jackson et al. (2020) Journal of Open Research Science
Detecting Jax...
Jax/JaxOpt are not installed on your system, unable to use Jax engine.
Detecting TensorFlow...
Unable to load both TensorFlow and tensorflow-probability. TensorFlow engine disabled.
No module named 'tensorflow'
False
Calculating weights for radars 0 and 1
Calculating weights for radars 1 and 0
Calculating weights for models...
Starting solver
rmsVR = 16.213679825765688
Total points: 412470
The max of w_init is 0.0
Total number of model points: 2342081
Nfeval | Jvel | Jmass | Jsmooth | Jbg | Jvort | Jmodel | Jpoint | Max w
0|4115.2916| 0.0000| 0.0000| 0.0000| 0.0000|11243.5276| 0.0000| 0.0000
---------------------------------------------------------------------------
KeyboardInterrupt Traceback (most recent call last)
Cell In[1], line 68
64 """
65 Perform the retrieval
66 """
67
---> 68 grid_tree = pydda.retrieval.get_dd_wind_field_nested(tree)
69
70 """
71 Plot the coarse grid output and finer grid output
File ~/work/PyDDA/PyDDA/pydda/retrieval/nesting.py:56, in get_dd_wind_field_nested(grid_tree, **kwargs)
54 elif len(grid_list) > 0:
55 my_kwargs = tree_attrs
---> 56 output_grids, output_parameters = get_dd_wind_field(grid_list, **my_kwargs)
57 output_parameters = output_parameters.__dict__
58 if in_parent is True:
File ~/work/PyDDA/PyDDA/pydda/retrieval/wind_retrieve.py:1587, in get_dd_wind_field(Grids, u_init, v_init, w_init, engine, **kwargs)
1580 w_init = new_grids[0]["w"].values.squeeze()
1582 if (
1583 engine.lower() == "scipy"
1584 or engine.lower() == "jax"
1585 or engine.lower() == "auglag"
1586 ):
-> 1587 return _get_dd_wind_field_scipy(
1588 new_grids, u_init, v_init, w_init, engine, **kwargs
1589 )
1590 elif engine.lower() == "tensorflow":
1591 return _get_dd_wind_field_tensorflow(
1592 new_grids, u_init, v_init, w_init, **kwargs
1593 )
File ~/work/PyDDA/PyDDA/pydda/retrieval/wind_retrieve.py:667, in _get_dd_wind_field_scipy(Grids, u_init, v_init, w_init, engine, points, vel_name, refl_field, u_back, v_back, z_back, frz, Co, Cm, Cx, Cy, Cz, Cb, Cv, Cmod, Cpoint, cvtol, gtol, Jveltol, Ut, Vt, low_pass_filter, mask_outside_opt, weights_obs, weights_model, weights_bg, max_iterations, mask_w_outside_opt, filter_type, filter_window, filter_order, leise_nstep, min_bca, max_bca, upper_bc, above, model_fields, output_cost_functions, roi, wind_tol, tolerance, const_boundary_cond, max_wind_mag, parallel)
665 parameters.print_out = False
666 if engine.lower() == "scipy":
--> 667 winds = fmin_l_bfgs_b(
668 J_function,
669 winds,
670 args=(parameters,),
671 maxiter=max_iterations,
672 pgtol=tolerance,
673 bounds=bounds,
674 fprime=grad_J,
675 callback=_vert_velocity_callback,
676 )
677 else:
679 def loss_and_gradient(x):
File /usr/share/miniconda/envs/pydda-docs/lib/python3.14/site-packages/scipy/optimize/_lbfgsb_py.py:259, in fmin_l_bfgs_b(func, x0, fprime, args, approx_grad, bounds, m, factr, pgtol, epsilon, maxfun, maxiter, callback, maxls)
249 callback = _wrap_callback(callback)
250 opts = {'maxcor': m,
251 'ftol': factr * np.finfo(float).eps,
252 'gtol': pgtol,
(...) 256 'callback': callback,
257 'maxls': maxls}
--> 259 res = _minimize_lbfgsb(fun, x0, args=args, jac=jac, bounds=bounds,
260 **opts)
261 d = {'grad': res['jac'],
262 'task': res['message'],
263 'funcalls': res['nfev'],
264 'nit': res['nit'],
265 'warnflag': res['status']}
266 f = res['fun']
File /usr/share/miniconda/envs/pydda-docs/lib/python3.14/site-packages/scipy/optimize/_lbfgsb_py.py:420, in _minimize_lbfgsb(fun, x0, args, jac, bounds, maxcor, ftol, gtol, eps, maxfun, maxiter, callback, maxls, finite_diff_rel_step, workers, **unknown_options)
412 _lbfgsb.setulb(m, x, low_bnd, upper_bnd, nbd, f, g, factr, pgtol, wa,
413 iwa, task, lsave, isave, dsave, maxls, ln_task)
415 if task[0] == 3:
416 # The minimization routine wants f and g at the current x.
417 # Note that interruptions due to maxfun are postponed
418 # until the completion of the current minimization iteration.
419 # Overwrite f and g:
--> 420 f, g = func_and_grad(x)
421 elif task[0] == 1:
422 # new iteration
423 n_iterations += 1
File /usr/share/miniconda/envs/pydda-docs/lib/python3.14/site-packages/scipy/optimize/_differentiable_functions.py:413, in ScalarFunction.fun_and_grad(self, x)
411 self._update_x(x)
412 self._update_fun()
--> 413 self._update_grad()
414 return self.f, self.g
File /usr/share/miniconda/envs/pydda-docs/lib/python3.14/site-packages/scipy/optimize/_differentiable_functions.py:375, in ScalarFunction._update_grad(self)
373 if self._orig_grad in FD_METHODS:
374 self._update_fun()
--> 375 self.g = self._wrapped_grad(self.x, f0=self.f)
376 self.g_updated = True
File /usr/share/miniconda/envs/pydda-docs/lib/python3.14/site-packages/scipy/optimize/_differentiable_functions.py:39, in _ScalarGradWrapper.__call__(self, x, f0, **kwds)
35 def __call__(self, x, f0=None, **kwds):
36 # Send a copy because the user may overwrite it.
37 # The user of this class might want `x` to remain unchanged.
38 if callable(self.grad):
---> 39 g = np.atleast_1d(self.grad(np.copy(x), *self.args))
40 elif self.grad in FD_METHODS:
41 g, dct = approx_derivative(
42 self.fun,
43 x,
44 f0=f0,
45 **self.finite_diff_options,
46 )
File ~/work/PyDDA/PyDDA/pydda/cost_functions/cost_functions.py:522, in grad_J(winds, parameters)
520 if parameters.parallel:
521 futures = []
--> 522 with ThreadPoolExecutor() as pool:
523 futures.append(
524 pool.submit(
525 _cost_functions_numpy.calculate_grad_radial_vel,
(...) 539 )
540 )
541 if parameters.Cm > 0:
File /usr/share/miniconda/envs/pydda-docs/lib/python3.14/concurrent/futures/_base.py:673, in Executor.__exit__(self, exc_type, exc_val, exc_tb)
672 def __exit__(self, exc_type, exc_val, exc_tb):
--> 673 self.shutdown(wait=True)
674 return False
File /usr/share/miniconda/envs/pydda-docs/lib/python3.14/concurrent/futures/thread.py:273, in ThreadPoolExecutor.shutdown(self, wait, cancel_futures)
271 if wait:
272 for t in self._threads:
--> 273 t.join()
File /usr/share/miniconda/envs/pydda-docs/lib/python3.14/threading.py:1133, in Thread.join(self, timeout)
1130 if timeout is not None:
1131 timeout = max(timeout, 0)
-> 1133 self._os_thread_handle.join(timeout)
KeyboardInterrupt: