Add Case38 validated redacted session and research summary

.gitattributes

@@ -103,3 +103,4 @@ cases/case40/session.science filter=lfs diff=lfs merge=lfs -text
 cases/case31/session.science filter=lfs diff=lfs merge=lfs -text
 cases/case37/session.science filter=lfs diff=lfs merge=lfs -text
 cases/case39/session.science filter=lfs diff=lfs merge=lfs -text
+cases/case38/session.science filter=lfs diff=lfs merge=lfs -text

cases/case38/README.md

@@ -0,0 +1,38 @@
+# Case 38: Heat-exposure modeling and coral resilience candidates
+
+## Research goal and case description
+
+This case asks whether public bleaching observations and independent NOAA temperature records identify Great Barrier Reef sites that bleach less or more than expected for their heat exposure. The analysis combines GCBD Reef Check observations with NOAA CoralTemp/Coral Reef Watch v3.1, models 407 site-years from 510 surveys at 153 sites (2003–2017), and evaluates site-held-out predictions, uncertainty and sensitivity. The adjusted model's held-out-site RMSE is 7.90 percentage points, compared with 11.24 for an intercept-only model. No site meets all prespecified candidate criteria; five coverage-eligible locations are proposed for additional sampling, not classified as resilient reefs.
+
+## Authors and attribution
+
+- Author and preferred attribution: Wyusyus.
+- Submitted by: Wyusyus.
+- Case ID: `case38`.
+- Session export date: 2026-10-19.
+
+## Session package
+
+- [`session.science`](session.science): retained research conversation and Notebook execution history, report, frozen specification, matched-survey and candidate tables, three scientific figures, and offline analysis bundle.
+
+This separate redacted derivative preserves the original export locally. Redaction replaced 462 local Windows user-profile prefixes in 21 captured Notebook PATH fields; scientific artifact bytes were unchanged. The package records the description with `kind: excluded`, and affected inventory and RO-Crate sizes and SHA-256 digests were recalculated. Targeted text screening found no remaining candidates within its coverage; large binary payloads were not exhaustively screened and image OCR was not performed.
+
+Native import succeeded on 2026-10-19 into a separate validation project using Open-Science Standalone 1.0.0 with local package-limit and Windows directory-publication fixes. The imported receipt retains 23,153 file rows and 88 Notebook runs; all eight scientific deliverables were byte-identical. Unpatched versions may still reject this large-history package. The uploaded file is unchanged since successful import validation.
+
+Submitted SHA-256: `ce68572db16724d621abc6d3e4d6616f0551f0109a47640a50c4b4074977aad5`.
+
+## Sharing scope
+
+| Use | Permission / conditions |
+|---|---|
+| AIPOCH team inspection and reproduction | Allowed for review of this submission, subject to third-party source terms. |
+| Case description or selected figures in the preprint | Confirm with the author before reuse; credit Wyusyus and respect third-party terms. |
+| Public distribution of the .science package and README | Requested by the submitter for this redacted submission, subject to third-party source terms. |
+
+Sources include the [Global Coral Bleaching Database v1](https://doi.org/10.6084/m9.figshare.17076290.v1) (CC0), its [associated paper](https://doi.org/10.1038/s41597-022-01121-y) (CC BY), and [NOAA Coral Reef Watch v3.1](https://coralreefwatch.noaa.gov/product/5km/methodology.php). Third-party materials retain their source terms; repository licensing does not expand their permissions.
+
+## Additional notes
+
+This is a historical observational screen, not evidence of thermal tolerance, adaptation, refuge status or current reef condition. Sparse repeated high-heat surveys, survey selection and timing, shared satellite pixels, local temperature mismatch and ecological composition limit inference. All simultaneous site intervals include zero; partial pooling reached the zero between-site-variance boundary. Absence of supported candidates does not establish that biological site differences are absent.
+
+The saved-input offline self-rerun reproduced 18 CSVs byte-for-byte; two figures were pixel-identical and a small raster difference in the third was disclosed. Native Reviewer returned four PASS checks after a retry, but this was not a complete independent computational audit. Native Replay remained unavailable because environment-lock and execution-capture requirements were incomplete; no successful Replay is claimed. One operator execution-recovery prompt and one review retry were used. Packaging fixes and successful import validation occurred later and do not change the scientific findings or constitute a scientific rerun.

cases/case38/session.science

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce68572db16724d621abc6d3e4d6616f0551f0109a47640a50c4b4074977aad5
+size 137236941