Add CYGNSS preprocessed scalar observation operator - #184
Draft
amfox37 wants to merge 21 commits into
Draft
Conversation
…y ig/jg Replace experiment-specific tile_index0 support-tile identity with globally stable tile_ig/tile_jg (M36 EASE-grid column/row). Both the coefficient operator cache (cygnss_preprocessed_obs.F90) and the observation reader (clsm_ensupd_read_obs.F90) now require schema_version = 0.4 and abort with a clear message on older files. Lookup is by tile_coord_lH%i_indg/j_indg instead of %f_num, so the coefficient product is portable across GEOSldas experiments with different tile counts or ordering. The missing-halo abort now reports ig/jg for easier diagnosis. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…a no-op The day-spanning loop in read_obs_cygnss_l1_scalar opens the previous UTC day's file for 00z-centered windows, but with the CYGNSS_operator preprocessor's daily-merge convention that file never contributes rows: each target day's file already has its own window 0 built from the prior day's 22:30-24:00 tail. Note this so the generic cross-day loop isn't mistaken for a bug or for missing data.
The schema-0.4 port already moved support-tile lookup to globally-stable ig/jg, but two other spots still trusted the preprocessor's own experiment-specific tile index: - read_obs_cygnss_l1_scalar (clsm_ensupd_read_obs.F90) assigned each observation to its owner tile via sp_nearest_tile_index0+1 directly as a local tile number. That index is tied to whatever domain the preprocessor itself was run against, so for a cropped/limited-domain experiment with a much smaller local tile count, nearly every value landed outside the valid range and got rejected as a "bad owner tile" -- 100% rejection in the worst case. Now resolved via sp_nearest_tile_ig/jg matched against this experiment's own tile_coord%i_indg/j_indg. - cygnss_preproc_load's per-observation cache (cygnss_preprocessed_obs.F90) had the same bug, plus the call site in clsm_ensupd_upd_routines.F90 was passing a full-domain tile number where a local-plus-halo index was expected. Fixed the cache the same way (now takes N_catlH/tile_coord_lH) and converted the call site via f2l(), the same full-to-local map already used for the generic FOV-based obs path one branch up. - cygnss_preproc_get_obs_pred aborted the whole run if any one observation's support tile fell outside the EnKF halo -- expected for observations near a cropped domain's edge. Now resolves all of an observation's support tiles up front and treats it as no-data rather than aborting if any are unresolvable (also drops the now-redundant per-ensemble-member repeated ig/jg lookup). Verified on both a limited-domain (CONUS box, Jul 2020) and the original full-global-domain (Nov 2019) experiment. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
cygnss_preproc_clear() existed but was only called internally to invalidate the cache before loading a new day; it was never wired into GridComp Finalize(), so the module-level cache arrays were missed by the recent memory-leak cleanup pass.
read_obs_cygnss_l1_scalar() guessed the coefficient product's EASE grid by substring-matching 'M36'/'M09' against the full file path. Since files are stored under Y<year>/M<month>/, any September file's path contains "M09" (the month directory), which clobbered a correct M36 match from the filename and aborted the run with a bogus grid mismatch. Schema-0.5 files already declare their grid explicitly via a 'grid' global attribute - read and trust that instead, keeping the old path/attribute substring guess only as a fallback for files without it, and dropping the full-path (tmpfname) search from that fallback since it's the source of the collision.
|
This PR is being prevented from merging because you have added one of our blocking labels: Contingent - DNA, Needs Lead Approval, Contingent -- Do Not Approve. You'll need to remove it before this PR can be merged. |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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.
This PR adds support for assimilating preprocessed CYGNSS L1 scalar observations.
Summary:
cygnss_preprocessed_obs.F90for reading/caching preprocessed daily CYGNSS files, including schema checks through version 0.5.Related PRs:
Care must be taken when develop is merged into the present branch after #186 is merged into develop.
There will be conflicts in GEOSldas_App/LDASsa_DEFAULT_inputs_ensupd.nml and GEOSlandassim_GridComp/clsm_ensupd_glob_param.F90
Testing:
cmake --build build-develop-20260520 --target install -j 4.