Surface statevector hookups and overhaul. by evan-greenbrg · Pull Request #871 · isofit/isofit

evan-greenbrg · 2026-02-18T01:41:19Z

Early stages, needs to be tested. Description also to be fleshed out.

This does two things:

Allow the user to specify state-vector optimization parameters directly in the surface.json file. This can be done by explicitly pointing to a surface model, or broadly including the relevant parameters.

I'll include more fleshed out specifics, but along the lines of:

  "SKY_GLINT": {
    "bounds": [
      0.0,
      10.0
    ],
    "init": 0.318,
    "prior_mean": 0.318,
    "prior_sigma": 0.001,
    "scale": 1
 },

These will propagate into the surface.mat file, which will in turn be read by the build_config script.

I updated how we read in surface statevector optimziation parameters. These now come from (if not specified) named tuples in the same format as the StatevectorConfigElement. Default StatevectorConfigElement are then used for the surface models that have non-rfl state-vector elements. Within this context, values specified in the surface.json act as "overrides" for the defaults.
I added a warnings section for check_config_validity(). Currently, this is only used for matching the non-reflectance surface statevector elements between .json and .mat files. Failing the check prints a warning:

WARNING:root:Configured non-reflectance surface statevector elements do not match .mat file.
Run will use configured values in the .json config.
Mismatching values:
SKY_GLINT: prior_mean
SUN_GLINT: init

evan-greenbrg · 2026-02-23T20:32:47Z

Replace get_single_element_by_name to get
Add commented text helpers with some of the old syntax to help with text searches to backtrack some of the syntax changes.
Review surface_thermal statevector terms.
Make sure the instrument statevector terms work.

davidraythompson · 2026-02-24T17:21:34Z

This looks good! @evan-greenbrg I like the way that folding glint priors into the .mat file makes it easier to keep the config file assumptions consistent with the .mat file. e.g. to avoid double bookkeeping glinty libraries with a physical glint model.

pgbrodrick · 2026-03-09T19:06:26Z

Ensure that setting the values is reasonably intuitive.

unbohn · 2026-03-10T00:00:55Z

Hitting an edge case here when initializing the forward model with an already existing main config file: if that file does not include the statevector key in the surface section, any updates in that respect from a freshly created mat file are not transferred to the main config, and not considered within the inversion.

Solution could be to implement a check if the surface mat file that is indicated in the main config provides the statevector key, and if not present in the main config as well, adopt those keys and values.

pgbrodrick · 2026-03-12T19:18:51Z

Discussed solution:

throw warning if there's a difference between .json input and .mat.

Surface statevector can come from surface.json. Overhaul internal surface statevector hookups Typo Breaking out RT, surface, instrument StateVectorConfig Debugging examples Logging bug Change get_config_single_element to get Added some additonal comments typo

…heck Backwards compatability Typo

github-actions · 2026-04-13T17:58:20Z

📊 Generated results:

URL: isofit/isofit-test-results#25
SHA: d2328bb

evan-greenbrg · 2026-04-14T17:01:21Z

This PR does not introduce any changes that by default impact the underlying physics or jacobian. This is supported by null deltas between dev branch solutions and this branch.

Reflectance:

Atmosphere:

pgbrodrick · 2026-04-15T02:38:29Z

Thanks for the tests @evan-greenbrg.

Given that sun & sky glint are now config terms, and given that the glint model requires (based on indexing) both of them to be populated in order to use, we should have a condition in the configuration that requires the mutual presence of both parameters if either is populated and we're in glint mode. I don't see it - am I missing it, or is this something we need to fill in?

Outside of this, I'm ready to merge.

evan-greenbrg · 2026-04-15T21:05:37Z

Thanks for the tests @evan-greenbrg.

Given that sun & sky glint are now config terms, and given that the glint model requires (based on indexing) both of them to be populated in order to use, we should have a condition in the configuration that requires the mutual presence of both parameters if either is populated and we're in glint mode. I don't see it - am I missing it, or is this something we need to fill in?

Outside of this, I'm ready to merge.

This is a good point. As it's currently written, the user could theoretically only provide one of the glint terms in the config, and the other config value would be pulled from the default. No matter if specified, the necessary glint config terms for the prior and scale values will be accessible.

We could add an explicit check that if one of the terms is specified custom within the config, the other would have to be as well. Surface statevector terms are treated independently from a config standpoint--analogous to the RT statevector terms.

pgbrodrick · 2026-04-15T22:31:43Z

Agreed - given this is an uncommon pathway, I'd be willing to merge as we flag to fix later - thoughts? Or would you prefer to fix first?

evan-greenbrg · 2026-04-20T18:29:46Z

Agreed - given this is an uncommon pathway, I'd be willing to merge as we flag to fix later - thoughts? Or would you prefer to fix first?

I'm conflicted on this. In the glint case, I think a check that raises some log if a Default config is used is a useful one. However, this PR would be nice to get in before the RT refactor as the config/state variable configs may change when removing RT_engines.

Barring any strong reservations, I would lean to consider merging this, then patching.

evan-greenbrg force-pushed the glint/statevector_config branch 2 times, most recently from 250ec24 to b2ad2a0 Compare February 18, 2026 21:09

evan-greenbrg marked this pull request as ready for review February 19, 2026 17:37

evan-greenbrg requested review from davidraythompson, pgbrodrick and unbohn March 3, 2026 00:09

evan-greenbrg force-pushed the glint/statevector_config branch from b2ad2a0 to ae0fb08 Compare March 9, 2026 15:07

brentwilder reviewed Mar 19, 2026

View reviewed changes

Comment thread isofit/utils/template_construction.py Outdated

brentwilder mentioned this pull request Mar 19, 2026

LUTSurface improvements #858

Draft

evan-greenbrg force-pushed the glint/statevector_config branch from 82defbc to f654b33 Compare March 23, 2026 16:52

evan-greenbrg requested a review from jammont March 23, 2026 19:23

evan-greenbrg added 2 commits April 13, 2026 08:48

Adding warnings to config check. Adding surface statevector content c…

d2328bb

…heck Backwards compatability Typo

evan-greenbrg force-pushed the glint/statevector_config branch from 7fe2088 to d2328bb Compare April 13, 2026 15:49

pgbrodrick mentioned this pull request Apr 13, 2026

871 isofit/isofit-test-results#25

Open

pgbrodrick merged commit b765f1a into isofit:dev Apr 20, 2026
27 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Surface statevector hookups and overhaul.#871

Surface statevector hookups and overhaul.#871
pgbrodrick merged 2 commits into
isofit:devfrom
evan-greenbrg:glint/statevector_config

evan-greenbrg commented Feb 18, 2026 •

edited

Loading

Uh oh!

evan-greenbrg commented Feb 23, 2026 •

edited

Loading

Uh oh!

davidraythompson commented Feb 24, 2026

Uh oh!

pgbrodrick commented Mar 9, 2026

Uh oh!

unbohn commented Mar 10, 2026

Uh oh!

pgbrodrick commented Mar 12, 2026

Uh oh!

Uh oh!

github-actions Bot commented Apr 13, 2026

Uh oh!

evan-greenbrg commented Apr 14, 2026 •

edited

Loading

Uh oh!

pgbrodrick commented Apr 15, 2026 •

edited

Loading

Uh oh!

evan-greenbrg commented Apr 15, 2026 •

edited

Loading

Uh oh!

pgbrodrick commented Apr 15, 2026

Uh oh!

evan-greenbrg commented Apr 20, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

evan-greenbrg commented Feb 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

evan-greenbrg commented Feb 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

davidraythompson commented Feb 24, 2026

Uh oh!

pgbrodrick commented Mar 9, 2026

Uh oh!

unbohn commented Mar 10, 2026

Uh oh!

pgbrodrick commented Mar 12, 2026

Uh oh!

Uh oh!

github-actions Bot commented Apr 13, 2026

Uh oh!

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

Uh oh!

Uh oh!

pgbrodrick commented Apr 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

evan-greenbrg commented Apr 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pgbrodrick commented Apr 15, 2026

Uh oh!

evan-greenbrg commented Apr 20, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

evan-greenbrg commented Feb 18, 2026 •

edited

Loading

evan-greenbrg commented Feb 23, 2026 •

edited

Loading

evan-greenbrg commented Apr 14, 2026 •

edited

Loading

pgbrodrick commented Apr 15, 2026 •

edited

Loading

evan-greenbrg commented Apr 15, 2026 •

edited

Loading