Coverage for C:\src\imod-python\imod\evaluate\budget.py: 89%
181 statements
« prev ^ index » next coverage.py v7.4.4, created at 2024-04-08 13:27 +0200
« prev ^ index » next coverage.py v7.4.4, created at 2024-04-08 13:27 +0200
1from typing import Tuple
3import dask
4import dask.array
5import numba
6import numpy as np
7import scipy.ndimage
8import xarray as xr
10DIM_Z = 0
11DIM_Y = 1
12DIM_X = 2
13LOWER = 3
14UPPER = 4
15FRONT = 5
16BACK = 6
17RIGHT = 7
18LEFT = 8
21def _outer_edge(da):
22 faces = np.full(da.shape, np.nan)
23 unique_ids = np.unique(da.values)
24 unique_ids = unique_ids[~np.isnan(unique_ids)]
25 # Number the faces by their id
26 for unique_id in unique_ids:
27 data = (da == unique_id).values
28 from_edge = ~scipy.ndimage.binary_dilation(~data)
29 is_edge = (data == 1) & (from_edge == 0)
30 faces[is_edge] = unique_id
31 return xr.DataArray(faces, da.coords, da.dims)
34@numba.njit
35def _face_indices(face, budgetzone):
36 nface = int(np.isfinite(face).sum())
37 shape = (nface, 9)
38 indices = np.zeros(shape, dtype=np.int32)
39 if nface == 0:
40 return indices
41 # Loop over cells
42 nlay, nrow, ncol = budgetzone.shape
43 count = 0
44 for k in range(nlay):
45 for i in range(nrow):
46 for j in range(ncol):
47 face_value = face[k, i, j]
48 if ~np.isnan(face_value):
49 # Store indices
50 indices[count, DIM_Z] = k
51 indices[count, DIM_Y] = i
52 indices[count, DIM_X] = j
53 # Default value: part of domain for edges
54 lower = front = right = upper = back = left = face_value
55 if k > 0:
56 upper = budgetzone[k - 1, i, j]
57 if k < (nlay - 1):
58 lower = budgetzone[k + 1, i, j]
59 if i > 0:
60 back = budgetzone[k, i - 1, j]
61 if i < (nrow - 1):
62 front = budgetzone[k, i + 1, j]
63 if j > 0:
64 left = budgetzone[k, i, j - 1]
65 if j < (ncol - 1):
66 right = budgetzone[k, i, j + 1]
68 # Test if cell is a control surface cell for the direction
69 if lower != face_value:
70 indices[count, LOWER] = 1
71 if upper != face_value:
72 indices[count, UPPER] = 1
73 if front != face_value:
74 indices[count, FRONT] = 1
75 if back != face_value:
76 indices[count, BACK] = 1
77 if right != face_value:
78 indices[count, RIGHT] = 1
79 if left != face_value:
80 indices[count, LEFT] = 1
82 # Incrementer counter
83 count += 1
84 return indices
87@numba.njit
88def _collect_flowfront(indices, flow):
89 result = np.zeros(flow.shape)
90 nface = indices.shape[0]
91 for count in range(nface):
92 k = indices[count, DIM_Z]
93 i = indices[count, DIM_Y]
94 j = indices[count, DIM_X]
95 if indices[count, FRONT]:
96 result[k, i, j] += flow[k, i, j]
97 if indices[count, BACK]:
98 result[k, i, j] -= flow[k, i - 1, j]
99 return result
102@numba.njit
103def _collect_flowlower(indices, flow):
104 result = np.zeros(flow.shape)
105 nface = indices.shape[0]
106 for count in range(nface):
107 k = indices[count, DIM_Z]
108 i = indices[count, DIM_Y]
109 j = indices[count, DIM_X]
110 if indices[count, LOWER]:
111 result[k, i, j] += flow[k, i, j]
112 if indices[count, UPPER]:
113 result[k, i, j] -= flow[k - 1, i, j]
114 return result
117@numba.njit
118def _collect_flowright(indices, flow):
119 result = np.zeros(flow.shape)
120 nface = indices.shape[0]
121 for count in range(nface):
122 k = indices[count, DIM_Z]
123 i = indices[count, DIM_Y]
124 j = indices[count, DIM_X]
125 if indices[count, RIGHT]:
126 result[k, i, j] += flow[k, i, j]
127 if indices[count, LEFT]:
128 result[k, i, j] -= flow[k, i, j - 1]
129 return result
132def delayed_collect(indices, front, lower, right):
133 indices = indices.compute()
134 result_front = dask.delayed(_collect_flowfront, nout=1)(indices, front.values)
135 result_lower = dask.delayed(_collect_flowlower, nout=1)(indices, lower.values)
136 result_right = dask.delayed(_collect_flowright, nout=1)(indices, right.values)
137 dask_front = dask.array.from_delayed(result_front, front.shape, dtype=np.float64)
138 dask_lower = dask.array.from_delayed(result_lower, lower.shape, dtype=np.float64)
139 dask_right = dask.array.from_delayed(result_right, right.shape, dtype=np.float64)
140 return dask_front, dask_lower, dask_right
143def facebudget(budgetzone, front=None, lower=None, right=None, netflow=True):
144 """
145 Computes net face flow into a control volume, as defined by ``budgetzone``.
147 Returns a three dimensional DataArray with in- and outgoing flows for every
148 cell that is located on the edge of the control volume, thereby calculating
149 the flow through the control surface of the control volume.
151 Front, lower, and right arguments refer to iMOD face flow budgets, in cubic
152 meters per day. In terms of flow direction these are defined as:
154 * ``front``: positive with ``y`` (negative with row index)
155 * ``lower``: positive with ``layer`` (positive with layer index)
156 * ``right``: negative with ``x`` (negative with column index)
158 Only a single face flow has to be defined, for example, if only vertical
159 fluxes (``lower``) are to be considered.
161 Note that you generally should not define a budgetzone that is only one cell
162 wide. In that case, flow will both enter and leave the cell, resulting in a
163 net flow of zero (given there are no boundary conditions present).
165 The output of this function defines ingoing flow as positive, and outgoing
166 flow as negative. The output is a 3D array with net flows for every control
167 surface cell. You can select specific parts of the control surface, for
168 example only the east-facing side of the control volume. Please refer to the
169 examples.
171 Parameters
172 ----------
173 budgetzone: xr.DataAray of floats
174 Array defining zones. Non-zones should be with a ``np.nan`` value.
175 Dimensions must be exactly ``("layer", "y", "x")``.
176 front: xr.DataArray of floats, optional
177 Dimensions must be exactly ``("layer", "y", "x")`` or
178 ``("time", "layer", "y", "x")``.
179 lower: xr.DataArray of floats, optional
180 Dimensions must be exactly ``("layer", "y", "x")`` or
181 ``("time", "layer", "y", "x")``.
182 right: xr.DataArray of floats, optional
183 Dimensions must be exactly ``("layer", "y", "x")`` or
184 ``("time", "layer", "y", "x")``.
185 netflow : bool, optional
186 Whether to split flows by direction (front, lower, right).
187 True: sum all flows. False: return individual directions.
189 Returns
190 -------
191 facebudget_front, facebudget_lower, face_budget_right : xr.DataArray of floats
192 Only returned if `netflow` is False.
193 facebudget_net : xr.DataArray of floats
194 Only returned if `netflow` is True.
196 Examples
197 --------
198 Load the face flows, and select the last time using index selection:
200 >>> import imod
201 >>> lower = imod.idf.open("bdgflf*.idf").isel(time=-1)
202 >>> right = imod.idf.open("bdgfrf*.idf").isel(time=-1)
203 >>> front = imod.idf.open("bdgfff*.idf").isel(time=-1)
205 Define the zone of interest, e.g. via rasterizing a shapefile:
207 >>> import geopandas as gpd
208 >>> gdf = gpd.read_file("zone_of_interest.shp")
209 >>> zone2D = imod.prepare.rasterize(gdf, like=lower.isel(layer=0))
211 Broadcast it to three dimensions:
213 >>> zone = xr.ones_like(flow) * zone2D
215 Compute net flow through the (control) surface of the budget zone:
217 >>> flow = imod.evaluate.facebudget(
218 >>> budgetzone=zone, front=front, lower=lower, right=right
219 >>> )
221 Or evaluate only horizontal fluxes:
223 >>> flow = imod.evaluate.facebudget(
224 >>> budgetzone=zone, front=front, right=right
225 >>> )
227 Extract the net flow, only on the right side of the zone, for example as
228 defined by x > 10000:
230 >>> netflow_right = flow.where(flow["x"] > 10_000.0).sum()
232 Extract the net flow for only the first four layers:
234 >>> netflow_layers = netflow_right.sel(layer=slice(1, 4)).sum()
236 Extract the net flow to the right of an arbitrary diagonal in ``x`` and
237 ``y`` is simple using the equation for a straight line:
239 >>> netflow_right_of_diagonal = flow.where(
240 >>> flow["y"] < (line_slope * flow["x"] + line_intercept)
241 >>> )
243 There are many ways to extract flows for a certain part of the zone of
244 interest. The most flexible one with regards to the ``x`` and ``y``
245 dimensions is by drawing a vector file, rasterizing it, and using it to
246 select with ``xarray.where()``.
248 To get the flows per direction, pass ``netflow=False``.
250 >>> flowfront, flowlower, flowright = imod.evaluate.facebudget(
251 >>> budgetzone=zone, front=front, lower=lower, right=right, netflow=False
252 >>> )
254 """
255 # Error handling
256 if front is None and lower is None and right is None:
257 raise ValueError("Atleast a single flow budget DataArray must be given")
258 if tuple(budgetzone.dims) != ("layer", "y", "x"):
259 raise ValueError('Dimensions of budgetzone must be exactly ("layer", "y", "x")')
260 shape = budgetzone.shape
261 for da, daname in zip((front, lower, right), ("front", "lower", "right")):
262 if da is not None:
263 dims = da.dims
264 coords = da.coords
265 if "time" in dims:
266 da_shape = da.shape[1:]
267 else:
268 da_shape = da.shape
269 if da_shape != shape:
270 raise ValueError(f"Shape of {daname} does not match budgetzone")
272 # Determine control surface
273 face = _outer_edge(budgetzone)
274 # Convert indices array to dask array, otherwise garbage collection gets
275 # rid of the array and we get a segfault.
276 indices = dask.array.from_array(_face_indices(face.values, budgetzone.values))
277 # Make sure it returns NaNs if no zones are defined.
278 if indices.size > 0:
279 # Create dummy arrays for skipped values, allocate just once
280 if front is None:
281 F = xr.full_like(budgetzone, 0.0, dtype=np.float64)
282 if lower is None:
283 L = xr.full_like(budgetzone, 0.0, dtype=np.float64)
284 if right is None:
285 R = xr.full_like(budgetzone, 0.0, dtype=np.float64)
287 results_front = []
288 results_lower = []
289 results_right = []
291 if "time" in dims:
292 for itime in range(front.coords["time"].size):
293 if front is not None:
294 F = front.isel(time=itime)
295 if lower is not None:
296 L = lower.isel(time=itime)
297 if right is not None:
298 R = right.isel(time=itime)
299 # collect dask arrays
300 dF, dL, dR = delayed_collect(indices, F, L, R)
301 # append
302 results_front.append(dF)
303 results_lower.append(dL)
304 results_right.append(dR)
306 # Concatenate over time dimension
307 dask_front = dask.array.stack(results_front, axis=0)
308 dask_lower = dask.array.stack(results_lower, axis=0)
309 dask_right = dask.array.stack(results_right, axis=0)
310 else:
311 if front is not None:
312 F = front
313 if lower is not None:
314 L = lower
315 if right is not None:
316 R = right
317 dask_front, dask_lower, dask_right = delayed_collect(indices, F, L, R)
318 else:
319 chunks = (1, *da_shape)
320 dask_front = dask.array.full(front.shape, np.nan, chunks=chunks)
321 dask_lower = dask.array.full(lower.shape, np.nan, chunks=chunks)
322 dask_right = dask.array.full(right.shape, np.nan, chunks=chunks)
324 if netflow:
325 return xr.DataArray(dask_front + dask_lower + dask_right, coords, dims)
326 else:
327 return (
328 xr.DataArray(dask_front, coords, dims),
329 xr.DataArray(dask_lower, coords, dims),
330 xr.DataArray(dask_right, coords, dims),
331 )
334def flow_velocity(
335 front: xr.DataArray,
336 lower: xr.DataArray,
337 right: xr.DataArray,
338 top_bot: xr.DataArray,
339 porosity: float = 0.3,
340) -> Tuple[xr.DataArray, xr.DataArray, xr.DataArray]:
341 """
342 Compute flow velocities (m/d) from budgets (m3/d).
344 Parameters
345 ----------
346 front: xr.DataArray of floats, optional
347 Dimensions must be exactly ``("layer", "y", "x")``.
348 lower: xr.DataArray of floats, optional
349 Dimensions must be exactly ``("layer", "y", "x")``.
350 right: xr.DataArray of floats, optional
351 Dimensions must be exactly ``("layer", "y", "x")``.
352 top_bot: xr.Dataset of floats, containing 'top', 'bot' and optionally
353 'dz' of layers.
354 Dimensions must be exactly ``("layer", "y", "x")``.
355 porosity: float or xr.DataArray of floats, optional (default 0.3)
356 If xr.DataArray, dimensions must be exactly ``("layer", "y", "x")``.
358 Returns
359 -------
360 vx, vy, vz: xr.DataArray of floats
361 Velocity components in x, y, z direction.
362 """
363 if "dz" not in top_bot:
364 top_bot["dz"] = top_bot["top"] - top_bot["bot"]
366 # cell side area (m2)
367 A_x = np.abs(top_bot.dz * right.dy)
368 A_y = np.abs(top_bot.dz * front.dx)
369 A_z = np.abs(lower.dx * lower.dy)
371 # Divide flux (m3/d) by area (m2) -> (m/d)
372 # Flip direction around for x (right)
373 return (
374 -right / (A_x * porosity),
375 front / (A_y * porosity),
376 lower / (A_z * porosity),
377 )