Coverage for src / CSET / operators / read.py: 90%

1# © Crown copyright, Met Office (2022-2025) and CSET contributors. 

2# 

3# Licensed under the Apache License, Version 2.0 (the "License"); 

4# you may not use this file except in compliance with the License. 

5# You may obtain a copy of the License at 

6# 

7# http://www.apache.org/licenses/LICENSE-2.0 

8# 

9# Unless required by applicable law or agreed to in writing, software 

10# distributed under the License is distributed on an "AS IS" BASIS, 

11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 

12# See the License for the specific language governing permissions and 

13# limitations under the License. 

14 

15"""Operators for reading various types of files from disk.""" 

16 

17import ast 

18import datetime 

19import functools 

20import glob 

21import itertools 

22import logging 

23from pathlib import Path 

24from typing import Literal 

25 

26import iris 

27import iris.coord_systems 

28import iris.coords 

29import iris.cube 

30import iris.exceptions 

31import iris.util 

32import numpy as np 

33from iris.analysis.cartography import rotate_pole, rotate_winds 

34 

35from CSET._common import iter_maybe 

36from CSET.operators._stash_to_lfric import STASH_TO_LFRIC 

37from CSET.operators._utils import ( 

38 get_cube_coordindex, 

39 get_cube_yxcoordname, 

40 is_spatialdim, 

41) 

42 

43 

44class NoDataError(FileNotFoundError): 

45 """Error that no data has been loaded.""" 

46 

47 

48def read_cube( 

49 file_paths: list[str] | str, 

50 constraint: iris.Constraint = None, 

51 model_names: list[str] | str | None = None, 

52 subarea_type: str = None, 

53 subarea_extent: list[float] = None, 

54 **kwargs, 

55) -> iris.cube.Cube: 

56 """Read a single cube from files. 

57 

58 Read operator that takes a path string (can include shell-style glob 

59 patterns), and loads the cube matching the constraint. If any paths point to 

60 directory, all the files contained within are loaded. 

61 

62 Ensemble data can also be loaded. If it has a realization coordinate 

63 already, it will be directly used. If not, it will have its member number 

64 guessed from the filename, based on one of several common patterns. For 

65 example the pattern *emXX*, where XX is the realization. 

66 

67 Deterministic data will be loaded with a realization of 0, allowing it to be 

68 processed in the same way as ensemble data. 

69 

70 Arguments 

71 --------- 

72 file_paths: str | list[str] 

73 Path or paths to where .pp/.nc files are located 

74 constraint: iris.Constraint | iris.ConstraintCombination, optional 

75 Constraints to filter data by. Defaults to unconstrained. 

76 model_names: str | list[str], optional 

77 Names of the models that correspond to respective paths in file_paths. 

78 subarea_type: "gridcells" | "modelrelative" | "realworld", optional 

79 Whether to constrain data by model relative coordinates or real world 

80 coordinates. 

81 subarea_extent: list, optional 

82 List of coordinates to constraint data by, in order lower latitude, 

83 upper latitude, lower longitude, upper longitude. 

84 

85 Returns 

86 ------- 

87 cubes: iris.cube.Cube 

88 Cube loaded 

89 

90 Raises 

91 ------ 

92 FileNotFoundError 

93 If the provided path does not exist 

94 ValueError 

95 If the constraint doesn't produce a single cube. 

96 """ 

97 cubes = read_cubes( 

98 file_paths=file_paths, 

99 constraint=constraint, 

100 model_names=model_names, 

101 subarea_type=subarea_type, 

102 subarea_extent=subarea_extent, 

103 ) 

104 # Check filtered cubes is a CubeList containing one cube. 

105 if len(cubes) == 1: 

106 return cubes[0] 

107 else: 

108 raise ValueError( 

109 f"Constraint doesn't produce single cube: {constraint}\n{cubes}" 

110 ) 

111 

112 

113def read_cubes( 

114 file_paths: list[str] | str, 

115 constraint: iris.Constraint | None = None, 

116 model_names: str | list[str] | None = None, 

117 subarea_type: str = None, 

118 subarea_extent: list = None, 

119 **kwargs, 

120) -> iris.cube.CubeList: 

121 """Read cubes from files. 

122 

123 Read operator that takes a path string (can include shell-style glob 

124 patterns), and loads the cubes matching the constraint. If any paths point 

125 to directory, all the files contained within are loaded. 

126 

127 Ensemble data can also be loaded. If it has a realization coordinate 

128 already, it will be directly used. If not, it will have its member number 

129 guessed from the filename, based on one of several common patterns. For 

130 example the pattern *emXX*, where XX is the realization. 

131 

132 Deterministic data will be loaded with a realization of 0, allowing it to be 

133 processed in the same way as ensemble data. 

134 

135 Data output by XIOS (such as LFRic) has its per-file metadata removed so 

136 that the cubes merge across files. 

137 

138 Arguments 

139 --------- 

140 file_paths: str | list[str] 

141 Path or paths to where .pp/.nc files are located. Can include globs. 

142 constraint: iris.Constraint | iris.ConstraintCombination, optional 

143 Constraints to filter data by. Defaults to unconstrained. 

144 model_names: str | list[str], optional 

145 Names of the models that correspond to respective paths in file_paths. 

146 subarea_type: str, optional 

147 Whether to constrain data by model relative coordinates or real world 

148 coordinates. 

149 subarea_extent: list[float], optional 

150 List of coordinates to constraint data by, in order lower latitude, 

151 upper latitude, lower longitude, upper longitude. 

152 

153 Returns 

154 ------- 

155 cubes: iris.cube.CubeList 

156 Cubes loaded after being merged and concatenated. 

157 

158 Raises 

159 ------ 

160 FileNotFoundError 

161 If the provided path does not exist 

162 """ 

163 # Get iterable of paths. Each path corresponds to 1 model. 

164 paths = iter_maybe(file_paths) 

165 model_names = iter_maybe(model_names) 

166 

167 # Check we have appropriate number of model names. 

168 if model_names != (None,) and len(model_names) != len(paths): 

169 raise ValueError( 

170 f"The number of model names ({len(model_names)}) should equal " 

171 f"the number of paths given ({len(paths)})." 

172 ) 

173 

174 # Load the data for each model into a CubeList per model. 

175 model_cubes = ( 

176 _load_model(path, name, constraint) 

177 for path, name in itertools.zip_longest(paths, model_names, fillvalue=None) 

178 ) 

179 

180 # Split out first model's cubes and mark it as the base for comparisons. 

181 cubes = next(model_cubes) 

182 for cube in cubes: 

183 # Use 1 to indicate True, as booleans can't be saved in NetCDF attributes. 

184 cube.attributes["cset_comparison_base"] = 1 

185 

186 # Load the rest of the models. 

187 cubes.extend(itertools.chain.from_iterable(model_cubes)) 

188 

189 # Unify time units so different case studies can merge. 

190 iris.util.unify_time_units(cubes) 

191 

192 # Select sub region. 

193 cubes = _cutout_cubes(cubes, subarea_type, subarea_extent) 

194 

195 # Merge and concatenate cubes now metadata has been fixed. 

196 cubes = cubes.merge() 

197 cubes = cubes.concatenate() 

198 

199 # Squeeze single valued coordinates into scalar coordinates. 

200 cubes = iris.cube.CubeList(iris.util.squeeze(cube) for cube in cubes) 

201 

202 # Ensure dimension coordinates are bounded. 

203 for cube in cubes: 

204 for dim_coord in cube.coords(dim_coords=True): 

205 # Iris can't guess the bounds of a scalar coordinate. 

206 if not dim_coord.has_bounds() and dim_coord.shape[0] > 1: 

207 dim_coord.guess_bounds() 

208 

209 logging.info("Loaded cubes: %s", cubes) 

210 if len(cubes) == 0: 

211 raise NoDataError("No cubes loaded, check your constraints!") 

212 return cubes 

213 

214 

215def _load_model( 

216 paths: str | list[str], 

217 model_name: str | None, 

218 constraint: iris.Constraint | None, 

219) -> iris.cube.CubeList: 

220 """Load a single model's data into a CubeList.""" 

221 input_files = _check_input_files(paths) 

222 # If unset, a constraint of None lets everything be loaded. 

223 logging.debug("Constraint: %s", constraint) 

224 cubes = iris.load(input_files, constraint, callback=_loading_callback) 

225 # Make the UM's winds consistent with LFRic. 

226 _fix_um_winds(cubes) 

227 

228 # Add model_name attribute to each cube to make it available at any further 

229 # step without needing to pass it as function parameter. 

230 if model_name is not None: 

231 for cube in cubes: 

232 cube.attributes["model_name"] = model_name 

233 return cubes 

234 

235 

236def _check_input_files(input_paths: str | list[str]) -> list[Path]: 

237 """Get an iterable of files to load, and check that they all exist. 

238 

239 Arguments 

240 --------- 

241 input_paths: list[str] 

242 List of paths to input files or directories. The path may itself contain 

243 glob patterns, but unlike in shells it will match directly first. 

244 

245 Returns 

246 ------- 

247 list[Path] 

248 A list of files to load. 

249 

250 Raises 

251 ------ 

252 FileNotFoundError: 

253 If the provided arguments don't resolve to at least one existing file. 

254 """ 

255 files = [] 

256 for raw_filename in iter_maybe(input_paths): 

257 # Match glob-like files first, if they exist. 

258 raw_path = Path(raw_filename) 

259 if raw_path.is_file(): 

260 files.append(raw_path) 

261 else: 

262 for input_path in glob.glob(raw_filename): 

263 # Convert string paths into Path objects. 

264 input_path = Path(input_path) 

265 # Get the list of files in the directory, or use it directly. 

266 if input_path.is_dir(): 

267 logging.debug("Checking directory '%s' for files", input_path) 

268 files.extend(p for p in input_path.iterdir() if p.is_file()) 

269 else: 

270 files.append(input_path) 

271 

272 files.sort() 

273 logging.info("Loading files:\n%s", "\n".join(str(path) for path in files)) 

274 if len(files) == 0: 

275 raise FileNotFoundError(f"No files found for {input_paths}") 

276 return files 

277 

278 

279def _cutout_cubes( 

280 cubes: iris.cube.CubeList, 

281 subarea_type: Literal["gridcells", "realworld", "modelrelative"] | None, 

282 subarea_extent: list[float, float, float, float], 

283): 

284 """Cut out a subarea from a CubeList.""" 

285 if subarea_type is None: 

286 logging.debug("Subarea selection is disabled.") 

287 return cubes 

288 

289 # If selected, cutout according to number of grid cells to trim from each edge. 

290 cutout_cubes = iris.cube.CubeList() 

291 # Find spatial coordinates 

292 for cube in cubes: 

293 # Find dimension coordinates. 

294 lat_name, lon_name = get_cube_yxcoordname(cube) 

295 

296 # Compute cutout based on number of cells to trim from edges. 

297 if subarea_type == "gridcells": 

298 logging.debug( 

299 "User requested LowerTrim: %s LeftTrim: %s UpperTrim: %s RightTrim: %s", 

300 subarea_extent[0], 

301 subarea_extent[1], 

302 subarea_extent[2], 

303 subarea_extent[3], 

304 ) 

305 lat_points = np.sort(cube.coord(lat_name).points) 

306 lon_points = np.sort(cube.coord(lon_name).points) 

307 # Define cutout region using user provided cell points. 

308 lats = [lat_points[subarea_extent[0]], lat_points[-subarea_extent[2] - 1]] 

309 lons = [lon_points[subarea_extent[1]], lon_points[-subarea_extent[3] - 1]] 

310 

311 # Compute cutout based on specified coordinate values. 

312 elif subarea_type == "realworld" or subarea_type == "modelrelative": 

313 # If not gridcells, cutout by requested geographic area, 

314 logging.debug( 

315 "User requested LLat: %s ULat: %s LLon: %s ULon: %s", 

316 subarea_extent[0], 

317 subarea_extent[1], 

318 subarea_extent[2], 

319 subarea_extent[3], 

320 ) 

321 # Define cutout region using user provided coordinates. 

322 lats = np.array(subarea_extent[0:2]) 

323 lons = np.array(subarea_extent[2:4]) 

324 # Ensure cutout longitudes are within +/- 180.0 bounds. 

325 while lons[0] < -180.0: 

326 lons += 360.0 

327 while lons[1] > 180.0: 

328 lons -= 360.0 

329 # If the coordinate system is rotated we convert coordinates into 

330 # model-relative coordinates to extract the appropriate cutout. 

331 coord_system = cube.coord(lat_name).coord_system 

332 if subarea_type == "realworld" and isinstance( 

333 coord_system, iris.coord_systems.RotatedGeogCS 

334 ): 

335 lons, lats = rotate_pole( 

336 lons, 

337 lats, 

338 pole_lon=coord_system.grid_north_pole_longitude, 

339 pole_lat=coord_system.grid_north_pole_latitude, 

340 ) 

341 else: 

342 raise ValueError("Unknown subarea_type:", subarea_type) 

343 

344 # Do cutout and add to cutout_cubes. 

345 intersection_args = {lat_name: lats, lon_name: lons} 

346 logging.debug("Cutting out coords: %s", intersection_args) 

347 try: 

348 cutout_cubes.append(cube.intersection(**intersection_args)) 

349 except IndexError as err: 

350 raise ValueError( 

351 "Region cutout error. Check and update SUBAREA_EXTENT." 

352 "Cutout region requested should be contained within data area. " 

353 "Also check if cutout region requested is smaller than input grid spacing." 

354 ) from err 

355 

356 return cutout_cubes 

357 

358 

359def _loading_callback(cube: iris.cube.Cube, field, filename: str) -> iris.cube.Cube: 

360 """Compose together the needed callbacks into a single function.""" 

361 # Most callbacks operate in-place, but save the cube when returned! 

362 _realization_callback(cube, field, filename) 

363 _um_normalise_callback(cube, field, filename) 

364 _lfric_normalise_callback(cube, field, filename) 

365 cube = _lfric_time_coord_fix_callback(cube, field, filename) 

366 _normalise_var0_varname(cube) 

367 _fix_spatial_coords_callback(cube) 

368 _fix_pressure_coord_callback(cube) 

369 _fix_um_radtime(cube) 

370 _fix_cell_methods(cube) 

371 cube = _convert_cube_units_callback(cube) 

372 cube = _grid_longitude_fix_callback(cube) 

373 _fix_lfric_cloud_base_altitude(cube) 

374 _proleptic_gregorian_fix(cube) 

375 _lfric_time_callback(cube) 

376 _lfric_forecast_period_callback(cube) 

377 return cube 

378 

379 

380def _realization_callback(cube, field, filename): 

381 """Give deterministic cubes a realization of 0. 

382 

383 This means they can be handled in the same way as ensembles through the rest 

384 of the code. 

385 """ 

386 # Only add if realization coordinate does not exist. 

387 if not cube.coords("realization"): 

388 cube.add_aux_coord( 

389 iris.coords.DimCoord(0, standard_name="realization", units="1") 

390 ) 

391 

392 

393@functools.lru_cache(None) 

394def _warn_once(msg): 

395 """Print a warning message, skipping recent duplicates.""" 

396 logging.warning(msg) 

397 

398 

399def _um_normalise_callback(cube: iris.cube.Cube, field, filename): 

400 """Normalise UM STASH variable long names to LFRic variable names. 

401 

402 Note standard names will remain associated with cubes where different. 

403 Long name will be used consistently in output filename and titles. 

404 """ 

405 # Convert STASH to LFRic variable name 

406 if "STASH" in cube.attributes: 

407 stash = cube.attributes["STASH"] 

408 try: 

409 (name, grid) = STASH_TO_LFRIC[str(stash)] 

410 cube.long_name = name 

411 except KeyError: 

412 # Don't change cubes with unknown stash codes. 

413 _warn_once( 

414 f"Unknown STASH code: {stash}. Please check file stash_to_lfric.py to update." 

415 ) 

416 

417 

418def _lfric_normalise_callback(cube: iris.cube.Cube, field, filename): 

419 """Normalise attributes that prevents LFRic cube from merging. 

420 

421 The uuid and timeStamp relate to the output file, as saved by XIOS, and has 

422 no relation to the data contained. These attributes are removed. 

423 

424 The um_stash_source is a list of STASH codes for when an LFRic field maps to 

425 multiple UM fields, however it can be encoded in any order. This attribute 

426 is sorted to prevent this. This attribute is only present in LFRic data that 

427 has been converted to look like UM data. 

428 """ 

429 # Remove unwanted attributes. 

430 cube.attributes.pop("timeStamp", None) 

431 cube.attributes.pop("uuid", None) 

432 cube.attributes.pop("name", None) 

433 cube.attributes.pop("source", None) 

434 cube.attributes.pop("analysis_source", None) 

435 cube.attributes.pop("history", None) 

436 

437 # Sort STASH code list. 

438 stash_list = cube.attributes.get("um_stash_source") 

439 if stash_list: 

440 # Parse the string as a list, sort, then re-encode as a string. 

441 cube.attributes["um_stash_source"] = str(sorted(ast.literal_eval(stash_list))) 

442 

443 

444def _lfric_time_coord_fix_callback( 

445 cube: iris.cube.Cube, field, filename 

446) -> iris.cube.Cube: 

447 """Ensure the time coordinate is a DimCoord rather than an AuxCoord. 

448 

449 The coordinate is converted and replaced if not. SLAMed LFRic data has this 

450 issue, though the coordinate satisfies all the properties for a DimCoord. 

451 Scalar time values are left as AuxCoords. 

452 """ 

453 # This issue seems to come from iris's handling of NetCDF files where time 

454 # always ends up as an AuxCoord. 

455 if cube.coords("time"): 

456 time_coord = cube.coord("time") 

457 if ( 

458 not isinstance(time_coord, iris.coords.DimCoord) 

459 and len(cube.coord_dims(time_coord)) == 1 

460 ): 

461 # Fudge the bounds to foil checking for strict monotonicity. 

462 if time_coord.has_bounds(): 462 ↛ 463line 462 didn't jump to line 463 because the condition on line 462 was never true

463 if (time_coord.bounds[-1][0] - time_coord.bounds[0][0]) < 1.0e-8: 

464 time_coord.bounds = [ 

465 [ 

466 time_coord.bounds[i][0] + 1.0e-8 * float(i), 

467 time_coord.bounds[i][1], 

468 ] 

469 for i in range(len(time_coord.bounds)) 

470 ] 

471 iris.util.promote_aux_coord_to_dim_coord(cube, time_coord) 

472 return cube 

473 

474 

475def _grid_longitude_fix_callback(cube: iris.cube.Cube) -> iris.cube.Cube: 

476 """Check grid_longitude coordinates are in the range -180 deg to 180 deg. 

477 

478 This is necessary if comparing two models with different conventions -- 

479 for example, models where the prime meridian is defined as 0 deg or 

480 360 deg. If not in the range -180 deg to 180 deg, we wrap the grid_longitude 

481 so that it falls in this range. Checks are for near-180 bounds given 

482 model data bounds may not extend exactly to 0. or 360. 

483 Input cubes on non-rotated grid coordinates are not impacted. 

484 """ 

485 try: 

486 y, x = get_cube_yxcoordname(cube) 

487 except ValueError: 

488 # Don't modify non-spatial cubes. 

489 return cube 

490 

491 long_coord = cube.coord(x) 

492 # Wrap longitudes if rotated pole coordinates 

493 coord_system = long_coord.coord_system 

494 if x == "grid_longitude" and isinstance( 

495 coord_system, iris.coord_systems.RotatedGeogCS 

496 ): 

497 long_points = long_coord.points.copy() 

498 long_centre = np.median(long_points) 

499 while long_centre < -175.0: 

500 long_centre += 360.0 

501 long_points += 360.0 

502 while long_centre >= 175.0: 

503 long_centre -= 360.0 

504 long_points -= 360.0 

505 long_coord.points = long_points 

506 

507 # Update coord bounds to be consistent with wrapping. 

508 if long_coord.has_bounds(): 

509 long_coord.bounds = None 

510 long_coord.guess_bounds() 

511 

512 return cube 

513 

514 

515def _fix_spatial_coords_callback(cube: iris.cube.Cube): 

516 """Check latitude and longitude coordinates name. 

517 

518 This is necessary as some models define their grid as on rotated 

519 'grid_latitude' and 'grid_longitude' coordinates while others define 

520 the grid on non-rotated 'latitude' and 'longitude'. 

521 Cube dimensions need to be made consistent to avoid recipe failures, 

522 particularly where comparing multiple input models with differing spatial 

523 coordinates. 

524 """ 

525 # Check if cube is spatial. 

526 if not is_spatialdim(cube): 

527 # Don't modify non-spatial cubes. 

528 return 

529 

530 # Get spatial coords and dimension index. 

531 y_name, x_name = get_cube_yxcoordname(cube) 

532 ny = get_cube_coordindex(cube, y_name) 

533 nx = get_cube_coordindex(cube, x_name) 

534 

535 # Remove spatial coords bounds if erroneous values detected. 

536 # Aims to catch some errors in input coord bounds by setting 

537 # invalid threshold of 10000.0 

538 if cube.coord(x_name).has_bounds() and cube.coord(y_name).has_bounds(): 

539 bx_max = np.max(np.abs(cube.coord(x_name).bounds)) 

540 by_max = np.max(np.abs(cube.coord(y_name).bounds)) 

541 if bx_max > 10000.0 or by_max > 10000.0: 

542 cube.coord(x_name).bounds = None 

543 cube.coord(y_name).bounds = None 

544 

545 # Translate [grid_latitude, grid_longitude] to an unrotated 1-d DimCoord 

546 # [latitude, longitude] for instances where rotated_pole=90.0 

547 if "grid_latitude" in [coord.name() for coord in cube.coords(dim_coords=True)]: 

548 coord_system = cube.coord("grid_latitude").coord_system 

549 pole_lat = getattr(coord_system, "grid_north_pole_latitude", None) 

550 if pole_lat == 90.0: 550 ↛ 551line 550 didn't jump to line 551 because the condition on line 550 was never true

551 lats = cube.coord("grid_latitude").points 

552 lons = cube.coord("grid_longitude").points 

553 

554 cube.remove_coord("grid_latitude") 

555 cube.add_dim_coord( 

556 iris.coords.DimCoord( 

557 lats, 

558 standard_name="latitude", 

559 var_name="latitude", 

560 units="degrees", 

561 coord_system=iris.coord_systems.GeogCS(6371229.0), 

562 circular=True, 

563 ), 

564 ny, 

565 ) 

566 y_name = "latitude" 

567 cube.remove_coord("grid_longitude") 

568 cube.add_dim_coord( 

569 iris.coords.DimCoord( 

570 lons, 

571 standard_name="longitude", 

572 var_name="longitude", 

573 units="degrees", 

574 coord_system=iris.coord_systems.GeogCS(6371229.0), 

575 circular=True, 

576 ), 

577 nx, 

578 ) 

579 x_name = "longitude" 

580 

581 # Create additional AuxCoord [grid_latitude, grid_longitude] with 

582 # rotated pole attributes for cases with [lat, lon] inputs 

583 if y_name in ["latitude"] and cube.coord(y_name).units in [ 

584 "degrees", 

585 "degrees_north", 

586 "degrees_south", 

587 ]: 

588 # Add grid_latitude AuxCoord 

589 if "grid_latitude" not in [ 589 ↛ 602line 589 didn't jump to line 602 because the condition on line 589 was always true

590 coord.name() for coord in cube.coords(dim_coords=False) 

591 ]: 

592 cube.add_aux_coord( 

593 iris.coords.AuxCoord( 

594 cube.coord(y_name).points, 

595 var_name="grid_latitude", 

596 units="degrees", 

597 ), 

598 ny, 

599 ) 

600 # Ensure input latitude DimCoord has CoordSystem 

601 # This attribute is sometimes lost on iris.save 

602 if not cube.coord(y_name).coord_system: 

603 cube.coord(y_name).coord_system = iris.coord_systems.GeogCS(6371229.0) 

604 

605 if x_name in ["longitude"] and cube.coord(x_name).units in [ 

606 "degrees", 

607 "degrees_west", 

608 "degrees_east", 

609 ]: 

610 # Add grid_longitude AuxCoord 

611 if "grid_longitude" not in [ 611 ↛ 625line 611 didn't jump to line 625 because the condition on line 611 was always true

612 coord.name() for coord in cube.coords(dim_coords=False) 

613 ]: 

614 cube.add_aux_coord( 

615 iris.coords.AuxCoord( 

616 cube.coord(x_name).points, 

617 var_name="grid_longitude", 

618 units="degrees", 

619 ), 

620 nx, 

621 ) 

622 

623 # Ensure input longitude DimCoord has CoordSystem 

624 # This attribute is sometimes lost on iris.save 

625 if not cube.coord(x_name).coord_system: 

626 cube.coord(x_name).coord_system = iris.coord_systems.GeogCS(6371229.0) 

627 

628 

629def _fix_pressure_coord_callback(cube: iris.cube.Cube): 

630 """Rename pressure coordinate to "pressure" if it exists and ensure hPa units. 

631 

632 This problem was raised because the AIFS model data from ECMWF 

633 defines the pressure coordinate with the name "pressure_level" rather 

634 than compliant CF coordinate names. 

635 

636 Additionally, set the units of pressure to be hPa to be consistent with the UM, 

637 and approach the coordinates in a unified way. 

638 """ 

639 for coord in cube.dim_coords: 

640 if coord.name() in ["pressure_level", "pressure_levels"]: 

641 coord.rename("pressure") 

642 

643 if coord.name() == "pressure": 

644 if str(cube.coord("pressure").units) != "hPa": 

645 cube.coord("pressure").convert_units("hPa") 

646 

647 

648def _fix_um_radtime(cube: iris.cube.Cube): 

649 """Move radiation diagnostics from timestamps which are output N minutes or seconds past every hour. 

650 

651 This callback does not have any effect for output diagnostics with 

652 timestamps exactly 00 or 30 minutes past the hour. Only radiation 

653 diagnostics are checked. 

654 Note this callback does not interpolate the data in time, only adjust 

655 timestamps to sit on the hour to enable time-to-time difference plotting 

656 with models which may output radiation data on the hour. 

657 """ 

658 try: 

659 if cube.attributes["STASH"] in [ 

660 "m01s01i207", 

661 "m01s01i208", 

662 "m01s02i205", 

663 "m01s02i201", 

664 "m01s01i207", 

665 "m01s02i207", 

666 "m01s01i235", 

667 ]: 

668 time_coord = cube.coord("time") 

669 

670 # Convert time points to datetime objects 

671 time_unit = time_coord.units 

672 time_points = time_unit.num2date(time_coord.points) 

673 # Skip if times don't need fixing. 

674 if time_points[0].minute == 0 and time_points[0].second == 0: 

675 return 

676 if time_points[0].minute == 30 and time_points[0].second == 0: 676 ↛ 677line 676 didn't jump to line 677 because the condition on line 676 was never true

677 return 

678 

679 # Subtract time difference from the hour from each time point 

680 n_minute = time_points[0].minute 

681 n_second = time_points[0].second 

682 # If times closer to next hour, compute difference to add on to following hour 

683 if n_minute > 30: 

684 n_minute = n_minute - 60 

685 # Compute new diagnostic time stamp 

686 new_time_points = ( 

687 time_points 

688 - datetime.timedelta(minutes=n_minute) 

689 - datetime.timedelta(seconds=n_second) 

690 ) 

691 

692 # Convert back to numeric values using the original time unit. 

693 new_time_values = time_unit.date2num(new_time_points) 

694 

695 # Replace the time coordinate with updated values. 

696 time_coord.points = new_time_values 

697 

698 # Recompute forecast_period with corrected values. 

699 if cube.coord("forecast_period"): 699 ↛ exitline 699 didn't return from function '_fix_um_radtime' because the condition on line 699 was always true

700 fcst_prd_points = cube.coord("forecast_period").points 

701 new_fcst_points = ( 

702 time_unit.num2date(fcst_prd_points) 

703 - datetime.timedelta(minutes=n_minute) 

704 - datetime.timedelta(seconds=n_second) 

705 ) 

706 cube.coord("forecast_period").points = time_unit.date2num( 

707 new_fcst_points 

708 ) 

709 except KeyError: 

710 pass 

711 

712 

713def _fix_cell_methods(cube: iris.cube.Cube): 

714 """To fix the assumed cell_methods in accumulation STASH from UM. 

715 

716 Lightning (m01s21i104), rainfall amount (m01s04i201, m01s05i201) and snowfall amount 

717 (m01s04i202, m01s05i202) in UM is being output as a time accumulation, 

718 over each hour (TAcc1hr), but input cubes show cell_methods as "mean". 

719 For UM and LFRic inputs to be compatible, we assume accumulated cell_methods are 

720 "sum". This callback changes "mean" cube attribute cell_method to "sum", 

721 enabling the cell_method constraint on reading to select correct input. 

722 """ 

723 # Shift "mean" cell_method to "sum" for selected UM inputs. 

724 if cube.attributes.get("STASH") in [ 

725 "m01s21i104", 

726 "m01s04i201", 

727 "m01s04i202", 

728 "m01s05i201", 

729 "m01s05i202", 

730 ]: 

731 # Check if input cell_method contains "mean" time-processing. 

732 if set(cm.method for cm in cube.cell_methods) == {"mean"}: 732 ↛ exitline 732 didn't return from function '_fix_cell_methods' because the condition on line 732 was always true

733 # Retrieve interval and any comment information. 

734 for cell_method in cube.cell_methods: 

735 interval_str = cell_method.intervals 

736 comment_str = cell_method.comments 

737 

738 # Remove input aggregation method. 

739 cube.cell_methods = () 

740 

741 # Replace "mean" with "sum" cell_method to indicate aggregation. 

742 cube.add_cell_method( 

743 iris.coords.CellMethod( 

744 method="sum", 

745 coords="time", 

746 intervals=interval_str, 

747 comments=comment_str, 

748 ) 

749 ) 

750 

751 

752def _convert_cube_units_callback(cube: iris.cube.Cube): 

753 """Adjust diagnostic units for specific variables. 

754 

755 Some precipitation diagnostics are output with unit kg m-2 s-1 and are 

756 converted here to mm hr-1. 

757 

758 Visibility diagnostics are converted here from m to km to improve output 

759 formatting. 

760 """ 

761 # Convert precipitation diagnostic units if required. 

762 varnames = filter(None, [cube.long_name, cube.standard_name, cube.var_name]) 

763 if any("surface_microphysical" in name for name in varnames): 

764 if cube.units == "kg m-2 s-1": 

765 logging.debug( 

766 "Converting precipitation rate units from kg m-2 s-1 to mm hr-1" 

767 ) 

768 # Convert from kg m-2 s-1 to mm s-1 assuming 1kg water = 1l water = 1dm^3 water. 

769 # This is a 1:1 conversion, so we just change the units. 

770 cube.units = "mm s-1" 

771 # Convert the units to per hour. 

772 cube.convert_units("mm hr-1") 

773 elif cube.units == "kg m-2": 773 ↛ 780line 773 didn't jump to line 780 because the condition on line 773 was always true

774 logging.debug("Converting precipitation amount units from kg m-2 to mm") 

775 # Convert from kg m-2 to mm assuming 1kg water = 1l water = 1dm^3 water. 

776 # This is a 1:1 conversion, so we just change the units. 

777 cube.units = "mm" 

778 

779 # Convert visibility diagnostic units if required. 

780 varnames = filter(None, [cube.long_name, cube.standard_name, cube.var_name]) 

781 if any("visibility" in name for name in varnames): 

782 if cube.units == "m": 782 ↛ 787line 782 didn't jump to line 787 because the condition on line 782 was always true

783 logging.debug("Converting visibility units m to km.") 

784 # Convert the units to km. 

785 cube.convert_units("km") 

786 

787 return cube 

788 

789 

790def _fix_lfric_cloud_base_altitude(cube: iris.cube.Cube): 

791 """Mask cloud_base_altitude diagnostic in regions with no cloud.""" 

792 varnames = filter(None, [cube.long_name, cube.standard_name, cube.var_name]) 

793 if any("cloud_base_altitude" in name for name in varnames): 

794 # Mask cube where set > 144kft to catch default 144.35695538058164 

795 cube.data = np.ma.masked_array(cube.data) 

796 cube.data[cube.data > 144.0] = np.ma.masked 

797 

798 

799def _fix_um_winds(cubes: iris.cube.CubeList): 

800 """To make winds from the UM consistent with those from LFRic. 

801 

802 Diagnostics of wind are not always consistent between the UM 

803 and LFric. Here, winds from the UM are adjusted to make them i 

804 consistent with LFRic. 

805 """ 

806 # Check whether we have components of the wind identified by STASH, 

807 # (so this will apply only to cubes from the UM), but not the 

808 # wind speed and calculate it if it is missing. Note that 

809 # this will be biased low in general because the components will mostly 

810 # be time averages. For simplicity, we do this only if there is just one 

811 # cube of a component. A more complicated approach would be to consider 

812 # the cell methods, but it may not be warranted. 

813 u_constr = iris.AttributeConstraint(STASH="m01s03i225") 

814 v_constr = iris.AttributeConstraint(STASH="m01s03i226") 

815 speed_constr = iris.AttributeConstraint(STASH="m01s03i227") 

816 try: 

817 if cubes.extract(u_constr) and cubes.extract(v_constr): 817 ↛ 818line 817 didn't jump to line 818 because the condition on line 817 was never true

818 if len(cubes.extract(u_constr)) == 1 and not cubes.extract(speed_constr): 

819 _add_wind_speed_um(cubes) 

820 # Convert winds in the UM to be relative to true east and true north. 

821 _convert_wind_true_dirn_um(cubes) 

822 except (KeyError, AttributeError): 

823 pass 

824 

825 

826def _add_wind_speed_um(cubes: iris.cube.CubeList): 

827 """Add windspeeds to cubes from the UM.""" 

828 wspd10 = ( 

829 cubes.extract_cube(iris.AttributeConstraint(STASH="m01s03i225"))[0] ** 2 

830 + cubes.extract_cube(iris.AttributeConstraint(STASH="m01s03i226"))[0] ** 2 

831 ) ** 0.5 

832 wspd10.attributes["STASH"] = "m01s03i227" 

833 wspd10.standard_name = "wind_speed" 

834 wspd10.long_name = "wind_speed_at_10m" 

835 cubes.append(wspd10) 

836 

837 

838def _convert_wind_true_dirn_um(cubes: iris.cube.CubeList): 

839 """To convert winds to true directions. 

840 

841 Convert from the components relative to the grid to true directions. 

842 This functionality only handles the simplest case. 

843 """ 

844 u_grids = cubes.extract(iris.AttributeConstraint(STASH="m01s03i225")) 

845 v_grids = cubes.extract(iris.AttributeConstraint(STASH="m01s03i226")) 

846 for u, v in zip(u_grids, v_grids, strict=True): 

847 true_u, true_v = rotate_winds(u, v, iris.coord_systems.GeogCS(6371229.0)) 

848 u.data = true_u.data 

849 v.data = true_v.data 

850 

851 

852def _normalise_var0_varname(cube: iris.cube.Cube): 

853 """Fix varnames for consistency to allow merging. 

854 

855 Some model data netCDF sometimes have a coordinate name end in 

856 "_0" etc, where duplicate coordinates of same name are defined but 

857 with different attributes. This can be inconsistently managed in 

858 different model inputs and can cause cubes to fail to merge. 

859 """ 

860 for coord in cube.coords(): 

861 if coord.var_name and coord.var_name.endswith("_0"): 

862 coord.var_name = coord.var_name.removesuffix("_0") 

863 if coord.var_name and coord.var_name.endswith("_1"): 

864 coord.var_name = coord.var_name.removesuffix("_1") 

865 if coord.var_name and coord.var_name.endswith("_2"): 865 ↛ 866line 865 didn't jump to line 866 because the condition on line 865 was never true

866 coord.var_name = coord.var_name.removesuffix("_2") 

867 if coord.var_name and coord.var_name.endswith("_3"): 867 ↛ 868line 867 didn't jump to line 868 because the condition on line 867 was never true

868 coord.var_name = coord.var_name.removesuffix("_3") 

869 

870 if cube.var_name and cube.var_name.endswith("_0"): 

871 cube.var_name = cube.var_name.removesuffix("_0") 

872 

873 

874def _proleptic_gregorian_fix(cube: iris.cube.Cube): 

875 """Convert the calendars of time units to use a standard calendar.""" 

876 try: 

877 time_coord = cube.coord("time")