Maintain a comparable record

JAXA documents a version-three problem that produced unintended sea-ice detections in parts of its SICE and OKID products for specified periods in 2021 and 2022. The correction history identifies a V3000-to-V3001 fix on March 8, 2022. This is a scoped processing issue, not evidence that every observation from Shikisai was invalid.

Imagine an original time-series exercise in which a step appears exactly when the processing version changes. Investigate whether the change reflects the environment, the method or both. Reprocessing earlier observations with a consistent method can help, but the revised files and their provenance should remain identifiable. Quietly mixing versions hides a possible explanation for the step.

In September 2026, EUMETSAT OSI SAF moved major sea-ice products to AMSR3 observations from GOSAT-GW. Selected AMSR2 products continued during an overlap period. That service transition does not itself establish that Shizuku stopped operating, and it does not turn GOSAT-GW into a GCOM spacecraft. The observing instrument, spacecraft program and downstream service are distinct identities.

A fictional service may retain the same familiar map title after changing its sensor input. A reader who checks only the title can miss the transition. Record the sensor and product identifier as well as the name, and inspect the stated continuity and delivery conditions instead of assuming they stayed identical.

Trace continuity across a record change

An overlap period allows measurements from two instruments to be compared under matched conditions. In an original simplified example, the same stable test scene yields 50 units from instrument A and 52 from instrument B. A later series switching from A to B could show a two-unit step even without an environmental change. One matched scene is not enough to establish a universal correction.

A stronger assessment compares many matched observations across relevant conditions and retains uncertainty. The aim is a defensible record of environmental change, not a perfectly smooth line produced by deleting inconvenient differences. Readers should be able to trace an apparent trend through the instruments, versions and validation that support it.