Update 3D field parameter with ensemble smoother with distance-based localization #12469
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
oddvarlia
changed the title
oddvarlia
force-pushed
the
distance_based_batch_update
branch
2 times, most recently
from
1554bb7 to
36554db
Compare
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #12469 +/- ##
==========================================
- Coverage 90.67% 90.67% -0.01%
==========================================
Files 429 429
Lines 29803 30073 +270
==========================================
+ Hits 27025 27268 +243
- Misses 2778 2805 +27
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
oddvarlia
force-pushed
the
distance_based_batch_update
branch
9 times, most recently
from
b52aa01 to
a9c9709
Compare
oddvarlia
force-pushed
the
distance_based_batch_update
branch
4 times, most recently
from
568900f to
3d044e4
Compare
oddvarlia
force-pushed
the
distance_based_batch_update
branch
7 times, most recently
from
e4b0c6f to
8f14437
Compare
oddvarlia
force-pushed
the
distance_based_batch_update
branch
from
8f14437 to
9f42875
Compare
oddvarlia
force-pushed
the
distance_based_batch_update
branch
12 times, most recently
from
799157d to
5471426
Compare
oddvarlia
force-pushed
the
distance_based_batch_update
branch
3 times, most recently
from
0bd1029 to
a464ab2
Compare
oddvarlia
force-pushed
the
distance_based_batch_update
branch
from
999445a to
01ed347
Compare
xjules
reviewed
Jan 19, 2026
oddvarlia
force-pushed
the
distance_based_batch_update
branch
2 times, most recently
from
4ee2df8 to
69bf521
Compare
… 3D fields Added functions: - update_3D_field_with_distance_esmda - _calc_max_number_of_layers_per_batch_for_distance_localization The function 'update_3D_field_with_distance_esmda' will update one 3D field parameter (or 2D field parameter when nz = 1) using the algorithm implemented in class DistanceESMDA in repo iterative_ensemble_smoother. This function requires that an object of DistanceESMDA is input. It will split the field parameter in batches when necessary due to memory constraints. The function '_calc_max_number_of_layers_per_batch_for_distance_localization' calculates the max size of a batch of field parameter values within the limit of available memory. update pyproject.toml to include gaussianfft to be used in tests
oddvarlia
force-pushed
the
distance_based_batch_update
branch
from
69bf521 to
6b1d6e2
Compare
xjules
reviewed
Jan 20, 2026
| ) | ||
|
|
||
|
|
||
| def calc_max_number_of_layers_per_batch_for_distance_localization( |
Contributor
There was a problem hiding this comment.
Maybe as an inspiration you can have a look at def _calculate_adaptive_batch_size ?
xjules
reviewed
Jan 20, 2026
| X_prior_3D = X_prior.reshape(nx, ny, nz, nreal) | ||
| # No update if no observations or responses | ||
| if Y is None or Y.shape[0] == 0: | ||
| # No update of the field parameters |
Contributor
There was a problem hiding this comment.
I think we can just return X_prior, yes.
Contributor
There was a problem hiding this comment.
you can also remove the comments, but keep the one above. # No update if no observations or responses
xjules
reviewed
Jan 20, 2026
| ) | ||
| ) # Use float64 | ||
| nlayer_per_batch = min(max_nlayers_per_batch, nz) | ||
| nbatch = int(nz / nlayer_per_batch) |
Contributor
There was a problem hiding this comment.
if nz==10 and nlayer_per_batch==7 then nbach = 1, is this correct?
xjules
reviewed
Jan 20, 2026
| nlayer_per_batch = min(max_nlayers_per_batch, nz) | ||
| nbatch = int(nz / nlayer_per_batch) | ||
|
|
||
| log_msg = "Minimum number of batches required due to " |
Contributor
There was a problem hiding this comment.
we can log nbatch and nlayer_per_batch and make the message concise
xjules
reviewed
Jan 20, 2026
Comment on lines
+739
to
+743
| log_msg = ( | ||
| f"Batch number: {batch_number}\n" | ||
| f"start layer : {start_layer_number}\n" | ||
| f"end layer : {end_layer_number - 1}" | ||
| ) |
xjules
reviewed
Jan 20, 2026
|
|
||
|
|
||
| @memory_usage_decorator(enabled=False) | ||
| def update_3D_field_with_distance_esmda( |
Contributor
There was a problem hiding this comment.
Let's double-check if iterative-ensemble-smoother would be a better place to host this code.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Function to update 3D field parameters with the method implemented in DistanceESMDA for ensemble smoother with
distance-based localization
Linked to: #12406
Approach
A 3D field parameter can often not be update in one go due to large memory requirements due to the Kalman gain matrix K and the localization matrix RHO. To handle this the field parameter can be split into subsets or batches. Each batch should be a set of layers of field parameter values defined in the field parameter grid (ertbox grid). Each batch must consists of a complete set of layers, not subset of layers of field parameters. The main reason for processing the update by batches of layers of field parameters is that the calculation of scaling factors (elements in the RHO matrix) can be done layer by layer due to the assumption that only lateral distance is of importance in the definition of influence of an observation, hence the calculation of RHO is done for one layer and then copied to all the other layers of the ertbox grid.