GEOPY-2799: Implement MVI with PDE inversion by domfournier · Pull Request #375 · MiraGeoscience/simpeg-drivers

domfournier · 2026-04-13T20:24:52Z

GEOPY-2799 - Implement MVI with PDE inversion

…ent unitest

Copilot

Pull request overview

This PR introduces support for a new “magnetic vector pde” workflow (forward + inversion) and integrates it into joint inversion drivers/tests, alongside updates to rotated-regularization handling and related plumbing.

Changes:

Added new magnetic_vector_pde driver package (forward/inversion/options) and new PDE UI JSON assets.
Updated joint drivers/tests to exercise MVI PDE in joint survey / cross-gradient contexts.
Refactored rotated-regularization operator handling (degrees vs radians) and adjusted nested/simulation factory logic for remanence mapping.

Reviewed changes

Copilot reviewed 42 out of 42 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
`simpeg_drivers/potential_fields/magnetic_vector_pde/*`	Adds PDE forward/inversion drivers and options for the new method.
`simpeg_drivers-assets/uijson/magnetic_vector_pde_*.ui.json`	Adds UI schemas for PDE forward/inversion runs.
`simpeg_drivers/components/models.py`	Extends vector-model logic to include PDE inversion type; adjusts gradient rotation handling.
`simpeg_drivers/components/factories/*`	Allows factories (survey/source/receiver/simulation/directives) to recognize “magnetic vector pde”.
`simpeg_drivers/joint/*`	Adjusts joint regularization/model handling to work with global mesh + PDE cases.
`simpeg_drivers/utils/regularization.py`	Updates rotated operator setup to accept degrees and convert internally.
`simpeg_drivers/utils/nested.py`	Ensures local tile mapping uses 3 components when remanence mapping is present.
`tests/run_tests/driver_mvi_pde_test.py`	Adds an integration-style test file for PDE forward+inversion.
`tests/run_tests/driver_joint_surveys_test.py`	Parametrizes joint surveys MVI test to include PDE driver/options.
`tests/run_tests/driver_joint_cross_gradient_test.py`	Adds a rotated-gradient joint cross-gradient test and tweaks inv run behavior.
`pyproject.toml` + conda lock files	Pins `mira-simpeg` and refreshes environment locks accordingly.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

simpeg_drivers/components/models.py

simpeg_drivers/potential_fields/magnetic_vector_pde/options.py

simpeg_drivers/components/factories/directives_factory.py

tests/run_tests/driver_mvi_pde_test.py

simpeg_drivers-assets/uijson/magnetic_vector_pde_inversion.ui.json

# Conflicts: # environments/py-3.12-linux-64-dev.conda.lock.yml # environments/py-3.12-linux-64.conda.lock.yml # environments/py-3.12-win-64-dev.conda.lock.yml # environments/py-3.12-win-64.conda.lock.yml # environments/py-3.13-linux-64-dev.conda.lock.yml # environments/py-3.13-linux-64.conda.lock.yml # environments/py-3.13-win-64-dev.conda.lock.yml # environments/py-3.13-win-64.conda.lock.yml # environments/py-3.14-linux-64-dev.conda.lock.yml # environments/py-3.14-linux-64.conda.lock.yml # environments/py-3.14-win-64-dev.conda.lock.yml # environments/py-3.14-win-64.conda.lock.yml # py-3.12.conda-lock.yml # py-3.13.conda-lock.yml # py-3.14.conda-lock.yml

This reverts commit b636fb3.

codecov · 2026-04-14T20:11:29Z

Codecov Report

❌ Patch coverage is 90.36697% with 21 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.21%. Comparing base (13c5823) to head (8904734).

Files with missing lines	Patch %	Lines
simpeg_drivers/driver.py	83.33%	5 Missing and 3 partials ⚠️
.../potential_fields/magnetic_vector_pde/inversion.py	81.48%	3 Missing and 2 partials ⚠️
...rs/potential_fields/magnetic_vector_pde/forward.py	70.00%	2 Missing and 1 partial ⚠️
...peg_drivers/components/factories/source_factory.py	77.77%	1 Missing and 1 partial ⚠️
simpeg_drivers/joint/driver.py	93.33%	1 Missing and 1 partial ⚠️
...drivers/components/factories/directives_factory.py	75.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop     #375      +/-   ##
===========================================
+ Coverage    90.05%   90.21%   +0.15%     
===========================================
  Files          127      131       +4     
  Lines         6368     6497     +129     
  Branches       795      809      +14     
===========================================
+ Hits          5735     5861     +126     
+ Misses         428      427       -1     
- Partials       205      209       +4

Files with missing lines	Coverage Δ
simpeg_drivers/__init__.py	`88.88% <ø> (ø)`
...g_drivers/components/factories/receiver_factory.py	`89.90% <100.00%> (ø)`
...peg_drivers/components/factories/simpeg_factory.py	`92.59% <ø> (ø)`
...drivers/components/factories/simulation_factory.py	`98.90% <100.00%> (+0.09%)`	⬆️
...peg_drivers/components/factories/survey_factory.py	`95.04% <100.00%> (ø)`
simpeg_drivers/components/models.py	`91.44% <100.00%> (ø)`
...impeg_drivers/joint/joint_cross_gradient/driver.py	`92.85% <100.00%> (+2.23%)`	⬆️
simpeg_drivers/joint/joint_surveys/driver.py	`93.05% <100.00%> (+0.86%)`	⬆️
simpeg_drivers/joint/joint_surveys/options.py	`92.30% <100.00%> (ø)`
...s/potential_fields/magnetic_vector_pde/__init__.py	`100.00% <100.00%> (ø)`
... and 11 more

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

benk-mira

Some questions/comments in the review. Also.. I wasn't expecting a whole new driver/options/ui.json combo. 1. Is the old method worth keeping around? (Will people want to choose between them on a case-by-case basis?) 2. Could this be written as an option in the mvi ui.json to switch from pde/integral equation?

benk-mira · 2026-04-14T21:21:37Z

simpeg_drivers/joint/joint_cross_gradient/driver.py

-        regularizations = super().get_regularization()
-        reg_list, multipliers = self._overload_regularization(regularizations)
+        # Trick the drivers by swapping the inversion_mesh and models
+        # such that the regularization uses the global mesh


This must be the fix for GEOPY-2656?

Yes, in big part. Also, because this new PDE Driver has a special regularization, we need to rely on each sub-inversion to generate their own reg. So the trick here is to collect all the models, interpolate onto the global mesh, then pass it back to the sub inversion to form their reg

simpeg_drivers/joint/driver.py

benk-mira · 2026-04-14T21:46:47Z

simpeg_drivers/potential_fields/magnetic_vector_pde/options.py

+    inducing_field_strength: float | FloatData
+    inducing_field_inclination: float | FloatData
+    inducing_field_declination: float | FloatData
+    models: VectorModelPDEOptions


Why no reference inclination/declination like in the regular mvi?

Because we currently don't have mechanics to enforce direction in Cartesian space. This new PDE formalulation doesn't allow to switch to spherical, so we can't allow it, for now.

benk-mira · 2026-04-14T21:53:19Z

simpeg_drivers/utils/meshes.py

    if any(
        k in inversion_type
-        for k in ["direct current", "induced polarization", "gravity", "magnetic"]
+        for k in ["direct current", "induced polarization", "gravity", "magnetic "]


why is the space needed? Seems like you could omit.. "magnetic scalar", "magnetic vector", and "magnetic vector pde" should all get captured without any confusion with "electromagnetic" as far as I can tell.

I don't think so, no. Because we do k in inversion_type, then "magnetic" in "electromagnetics" is True, which would take all of the EM methods. The space prevents that.

domfournier

Fair point that there is a lot of overlap, but there is enough for us to keep both around, for now. Main difference is that the PDE can't do reference angles, and has a special VectorAmplitude reg.
I think it would be possible to do reference angles in Cartesian space, but I ll need more time to develop. For now it's just simpler to have a second.

domfournier · 2026-04-15T03:43:39Z

simpeg_drivers/joint/joint_cross_gradient/driver.py

-        regularizations = super().get_regularization()
-        reg_list, multipliers = self._overload_regularization(regularizations)
+        # Trick the drivers by swapping the inversion_mesh and models
+        # such that the regularization uses the global mesh


Yes, in big part. Also, because this new PDE Driver has a special regularization, we need to rely on each sub-inversion to generate their own reg. So the trick here is to collect all the models, interpolate onto the global mesh, then pass it back to the sub inversion to form their reg

domfournier · 2026-04-15T03:45:12Z

simpeg_drivers/potential_fields/magnetic_vector_pde/options.py

+    inducing_field_strength: float | FloatData
+    inducing_field_inclination: float | FloatData
+    inducing_field_declination: float | FloatData
+    models: VectorModelPDEOptions


Because we currently don't have mechanics to enforce direction in Cartesian space. This new PDE formalulation doesn't allow to switch to spherical, so we can't allow it, for now.

domfournier · 2026-04-15T03:47:21Z

simpeg_drivers/utils/meshes.py

    if any(
        k in inversion_type
-        for k in ["direct current", "induced polarization", "gravity", "magnetic"]
+        for k in ["direct current", "induced polarization", "gravity", "magnetic "]


I don't think so, no. Because we do k in inversion_type, then "magnetic" in "electromagnetics" is True, which would take all of the EM methods. The space prevents that.

Co-authored-by: benk-mira <81254271+benk-mira@users.noreply.github.com>

domfournier added 20 commits April 9, 2026 14:09

Sprinkle "magnetic vector pde" throughout

7a9086e

MOve setter of mapping closer to property

79d9e7f

Refactor modification of reg into function

83010d2

Fix typo

d5f50fb

First run through

afa115f

Add back starting inc/dec for mvi pde

2dc0a17

Fix issue with res model not trimmed

24232bf

Fix tiling of active cells

b8f7756

Remove unused function

b661e96

Minor fixes

f41af0d

Make sure of VectorAmplitude for mvi pde

bf184b3

Tweek unitest

48ac908

Fixes after running through joint cross

4200218

Pass the reg models to sub-drivers. Keep reg from individual drivers

52f9528

Deal with switch in MVI and cross-gradient

f5f0879

Deal with non-trimmed models. Refactor get_regularization for MVI PDE

28903c5

Temp lock on simpeg@2799

7402995

Keep gradient_direction and dip in degrees until usage

e2ea8de

Augment unitest for gradient_rotation passed from the sub-drivers

10f698c

Fix joint survey to take into account driver.get_regularization. Augm…

3d00034

…ent unitest

Copilot AI review requested due to automatic review settings April 13, 2026 20:24

github-actions bot changed the title ~~GEOPY-2799~~ GEOPY-2799: Implement MVI with PDE inversion Apr 13, 2026

Copilot started reviewing on behalf of domfournier April 13, 2026 20:25 View session

Copilot AI reviewed Apr 13, 2026

View reviewed changes

benk-mira and others added 6 commits April 13, 2026 14:50

Merge branch 'develop' into GEOPY-2799

889be72

Fix pde test

0fe360f

Fix target for mvi pde

e825c7e

Copilot fixes

5caf071

Use directive on mvi pde to apply bounds

b636fb3

domfournier added 2 commits April 14, 2026 11:54

Revert "Use directive on mvi pde to apply bounds"

e12e395

This reverts commit b636fb3.

Add expection to not expand mesh size model

99b34dd

benk-mira requested changes Apr 14, 2026

View reviewed changes

domfournier commented Apr 15, 2026

View reviewed changes

Update simpeg_drivers/joint/driver.py

8904734

Co-authored-by: benk-mira <81254271+benk-mira@users.noreply.github.com>

Conversation

domfournier commented Apr 13, 2026 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

codecov bot commented Apr 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

benk-mira left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

domfournier left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

domfournier commented Apr 13, 2026 •

edited by github-actions bot

Loading

codecov bot commented Apr 14, 2026 •

edited

Loading