Edited sky glint prior, added indexing in glint model, edited bands in fit_params#718
Conversation
|
@evan-greenbrg Looks good! Once minor issues (see above) are resolved, I'll mark this ready for merge. |
|
Sorry I missed this today….generally looks good to me. I am curious if the updated algebraic inversion helps mitigate the need for the simple guess, but everything else looks great. Thanks especially for handling the sun/sky glint indexing |
…or fit_params Updates to AL and sky glint prior Edits to sky glint prior Editing glint fit parameters and adding indexing Simplifying fit_params wl choice Fixing typoes in surface_glint_model
138c7de to
1989923
Compare
Need to test the best first guess seeding with the glint model. Generally however, I reverted the default AOE first guess seeding. Algebraic performs well for the non-glint model and it's faster. If the glint model needs to use a different seeding, we can always change as needed. |
|
@pgbrodrick Please go ahead and merge once you have a minute. |
This PR:
Formalizes edits to the sky glint prior. After testing, the simplest solution to negative sky glint values is to use a tightly constrained prior. This fits with intuition that the diffuse sky glint term should be spatially smooth.
Glint-corrected solutions with EMIT images showed discrete brightness variations across the images s.t. some regions appeared brighter in the visible than others. A possible fix is to use longer wavelengths when available to set the initial reflectance guess.
Added sun and sky glint indexing variables to move away from nameless integer indexing throughout the glint model.
Results from an EMIT granule (NOTE glint-corrected uses the 1st version of the emulator, multicomponent uses stable sRTMnet)
