Tensor interpolation for non-stationary geostat fields #641
Conversation
…tensor and changed the config_func_str in pst_from to use it instead of value interpolation. this avoids issues with interpolating angles and I believe is more propa.
…rything 2d for easier comprehension
…is messed up, probably generate_from_files
…d examples. normalize weights in _moving_average_field still open question.
briochh
left a comment
briochh
left a comment
There was a problem hiding this comment.
A lot of new stuff here @wkitlasten, it may sit better in its own module at some point (or in ppu). For now I would support bringing it into pyemu under a prototypes sub-module while under development and getting tests sorted out. I think you could create a new prototypes dir under pyemu:
prototypes/
|--tests/
| nsaf_tests.py
|--src/
| |--nsaf/
| nsaf_utils.py
| nsaf_helpers.py
|--examples/
| |--Wairau/
| |--HBHP/
| pstfrom_mf6_ppu_tensor.ipynb
we can then add some specific GitHub Actions triggers (or just trigger requests) to run a specific suite over these without compromising the existing tests
| # self.extra_py_imports.append("pypestutils") | ||
| print(config_df_filename) | ||
| config_func_str = "pyemu.utils.pp_utils.apply_ppu_hyperpars('{0}')".\ | ||
| config_func_str = "pyemu.utils.pp_utils.apply_ppu_hyperpars_tensor('{0}')".\ |
There was a problem hiding this comment.
breaks current non-stationary hyper par implementation... I think best to option "tensor" (at least until it is proven)
There was a problem hiding this comment.
wondering if these nsaf funcs would fit better within pypestutils or a stand alone library (or submodule here). For now might be good to move into a prototypes submodule here. That way you we can keep reasonably contained (incl. dependancies) while refined and testing funcs better targeted.
| import os | ||
| import matplotlib as mpl | ||
|
|
||
| mpl.use('Tkagg') |
There was a problem hiding this comment.
this will cause failure on some CI platform (on import)
| "import numpy as np\n", | ||
| "import pandas as pd\n", | ||
| "import matplotlib\n", | ||
| "matplotlib.use('Qt5Agg')\n", |
There was a problem hiding this comment.
CI might not like this line
| "shutil.copy(r'C:\\bin\\modflow\\mf6.6.2_win64\\bin\\mf6.exe', os.path.join(tmp_model_ws,'mf6.exe'))\n", | ||
| "shutil.copy(r'C:\\bin\\pestpp\\pestpp-ies.exe', os.path.join(tmp_model_ws,'pestpp-ies.exe'))\n", |
There was a problem hiding this comment.
Will need to be smarter here with bin/exe definition for x-platform and CI
There was a problem hiding this comment.
much safer to clean notebook outputs before adding to git!
3 participants
Some utilities to perform log-Euclidean interpolation of anisotropic variance tensors from conceptual points to a regular grid. Also performs spatial convolution to create non-stationary geostatistical fields from variance tensors. Uses pypestutils where possible, with fallback to python implementation of moving average if needed (e.g., pypestutils fieldgen2d_sva does not accept external iid fields).