Find your next learning step
Correcting a remote-sensing product is only the first repair. Derived maps, models, reports and decisions may still contain the earlier pixels or metadata. Amendment propagation maps that dependency chain, decides which downstream work needs reanalysis, and gives every user a version-specific citation. The goal is not to rewrite history; it is to make current evidence findable while preserving exactly which product supported each earlier conclusion.
Open the chapter contents · Remote Sensing root · Gold Standard of Education · How X Works
Chapter contents
Choose the route closest to the decision, or read from Chapter 1. The chapters remain open; these grouped routes are a navigable index.
Route 1 · Define the decision
Route 2 · Work the ledgers
Route 3 · Transfer the method
Route 4 · Check limits and evidence
Route 5 · Ask and continue
CHAPTER 1 OF 12 · DEFINE THE DECISION
1. Treat an amendment as a dependency event
Describe the changed product version, scenes, dates, bands, strata, error mechanism and expected direction of change. Then identify consumers: mosaics, indices, classifiers, trend series, publications, dashboards and operational decisions.
A corrected source does not automatically correct a derived product. Propagation requires provenance or a reproducible inventory. When lineage is missing, uncertainty about use becomes part of the incident.
CHAPTER 2 OF 12 · DEFINE THE DECISION
2. Separate replacement, reanalysis and notice
Replacement supplies corrected assets. Reanalysis reruns a downstream method. A notice explains impact when rerunning is impossible, unnecessary or would change a historical record. Many chains need all three at different nodes.
Preserve the original product identifier, acquisition time and processing version. A current recommendation should link backward; an old publication should not be edited to imply it used data released later.
CHAPTER 3 OF 12 · WORK THE LEDGERS
3. Build an amendment-propagation register
A coastal land-cover amendment corrects aquaculture boundaries in 42 scenes.
| Downstream object | Dependency | Decision consequence | Reanalysis status | User action |
|---|---|---|---|---|
| Annual coastal mosaic | Direct pixels | High | Required | Replace release |
| Habitat trend model | Mosaic areas | High | Queued | Pause new claims |
| Local planning dashboard | Class totals | Medium | Required | Display notice |
| Archived field report | Figure only | Low | Not rerun | Attach amendment |
| Mangrove-only study | Unaffected stratum | Low | Documented exemption | No numerical change |
The register names an owner and version for every downstream node so acknowledgement does not substitute for repair.
CHAPTER 4 OF 12 · WORK THE LEDGERS
4. Work a reanalysis priority score
The programme scores consequence from 1–4, estimated numerical sensitivity from 1–4 and reproducibility readiness from 0–2. Priority equals consequence plus sensitivity plus readiness.
| Object | Consequence | Sensitivity | Readiness | Priority | Decision |
|---|---|---|---|---|---|
| Coastal mosaic | 4 | 4 | 2 | 10 | Immediate |
| Habitat trend model | 4 | 3 | 2 | 9 | Immediate |
| Planning dashboard | 3 | 3 | 2 | 8 | Next release |
| Archived field report | 1 | 1 | 0 | 2 | Notice only |
| Mangrove-only study | 1 | 0 | 2 | 3 | Exemption review |
The score is a transparent local rule, not a universal risk formula. The team can override it only with a dated reason.
CHAPTER 5 OF 12 · WORK THE LEDGERS
5. Trace indirect and partial dependencies
A model may depend on the amended product through training labels, a calibration mask or a summary table rather than a visible raster. Search code, manifests, DOIs, filenames, checksums and workflow logs.
Where only part of a time series changed, rerun both the affected interval and the trend diagnostics. A local correction can alter breakpoints or slopes even if the full-period mean barely moves.
CHAPTER 6 OF 12 · TRANSFER THE METHOD
6. Govern citations by version
Give major versions distinct, resolvable identifiers and link them with previous-version and new-version relations. For minor metadata correction, update the record while preserving a change log and access date.
A citation should identify title, producer, version or collection, release date, persistent identifier and accessed asset when those facts exist. Derived products should name the exact upstream version, not simply Landsat data or latest map.
CHAPTER 7 OF 12 · TRANSFER THE METHOD
7. Changed-context transfer: rapid flood products
Transfer the method to flood maps corrected after canopy-obscured inundation is found. Emergency maps, exposure counts and relief-routing dashboards depend on the released layers in different ways.
The current map is replaced, exposure counts are recomputed and historical response logs keep their original version with an attached amendment. The team does not claim that responders possessed the corrected extent during the event.
CHAPTER 8 OF 12 · TRANSFER THE METHOD
8. Connect repair, reclosure and collection history
Start with reopened-stratum repair and historical product amendment. Use verification-window reopening and rolling debt reassessment to recover why the amendment was issued.
Return to delayed-label backfill and verification-window closure when downstream work used evidence that arrived after the original release.
CHAPTER 9 OF 12 · CHECK LIMITS AND EVIDENCE
9. Limits and common misconceptions
A new source version does not make every earlier conclusion wrong. Impact depends on spatial, temporal and analytical dependence. Conversely, a small changed area can matter greatly if it controls a threshold or vulnerable group.
Latest is not a sufficient citation. Version governance cannot reconstruct lineage that was never recorded, and a DOI alone does not prove the correct files were used. Preserve checksums or manifests where practical.
CHAPTER 10 OF 12 · CHECK LIMITS AND EVIDENCE
10. Practical evidence and communication checks
Archive amendment scope, source and replacement IDs, scene list, checksums, error mechanism, changed metrics, dependency search, affected owners, reanalysis plan, exemptions, notices and closure criteria.
For citations, record persistent identifier, version relation, release and access dates, landing page, asset manifest, recommended citation and wording that distinguishes original publication context from current evidence.
CHAPTER 11 OF 12 · ASK AND CONTINUE
11. Frequently asked questions
Must every derivative be rerun?
No. Use documented dependency and consequence checks; unaffected derivatives can receive a recorded exemption.
Should an old product disappear?
Usually preserve its manifest and version record, even when access is restricted or the current landing page recommends the replacement.
Can a citation point to the collection rather than a scene?
Yes when the analysis truly used that collection-level object, but record the exact version and asset subset needed to reproduce the work.
CHAPTER 12 OF 12 · ASK AND CONTINUE
12. Sources and next reading
USGS states on its Landsat Collection 2 page that selected scenes may be reprocessed, product identifiers then reflect the latest processing date, and users should consider replacing downloaded data and derived products. The Collection 2 Reprocessing Events page records affected events, while DataCite’s versioning guidance recommends updating DOI metadata and assigning a new DOI for major changes with explicit version relations. The propagation register and scores are original teaching examples.
Continue through the Remote Sensing root, the Gold Standard of Education and How X Works.
Top · Contents · Previous · How X Works