Keep a long record comparable

A long record can reveal change only if the measuring system’s changes are understood. Instruments age; calibration and radiative-transfer models improve; processing rules evolve. Reprocessing an old observation can therefore produce a better estimate without changing the time when the original radiation was measured.

The calibration figure compares Version 8 near-infrared, Version 9 near-infrared and Version 9 thermal-only total columns for daytime land observations between 60°S and 60°N. Gray bars mark annual calibration and decontamination intervals. In the lower panel, the near-infrared products are differenced against the thermal reference, helping reveal discontinuities near those interventions.

Deeter and colleagues found that revising the near-infrared calibration reduced these discontinuities. That supports improved temporal consistency; it does not imply the thermal reference is perfect or every short-term fluctuation is an artifact. Instrument history helps distinguish a plausible atmospheric event from an effect introduced by the observing system.

Imagine an original index example: older processing reports 100 for an early year, while newer processing reports 90 for a later year. It is tempting to call that a ten-percent decline. But if the early year becomes 95 when processed with the same newer method, the comparable change is (90 − 95) / 95, about a 5.3-percent decline. These are invented index values, not MOPITT observations.

Consistent processing is necessary but does not settle every problem. The earlier lesson showed how cloud screening changes the accepted sample. A robust analysis also checks coverage, seasonal selection, sensitivity and uncertainty. Removing inconvenient observations after seeing the result can itself introduce bias.

Product maturity also matters. The Version 9 documentation labels provisional files that lack the full calibration information as beta products and directs quantitative users to the standard archive. The mission team’s version history now includes Version 10. The Version 9 study remains useful as a worked example of validation; it should not be presented as the latest product or silently generalized to every later release.

Compare both years with the same processing

The instrument team places the end of measurements at February 1, 2025, while NCAR lists the archived observation interval through January 31. Those describe the end-of-January boundary. NASA separately records MOPITT being powered off on April 9, 2025 to conserve Terra’s remaining power. Last observation, power-off and archive availability are distinct events.

The shutdown did not erase the collected record. Researchers can still reprocess and compare historical observations, provided the observation dates, product versions and uncertainties remain explicit. A new dataset release after shutdown is not evidence that the instrument resumed observing.

For your own analysis plan, name the gas and product, select a consistent version and period, inspect coverage, compare through the appropriate measurement response, and retain the source files and processing choices. MOPITT’s legacy is both a long atmospheric record and a practical lesson in how scientific claims stay traceable as methods improve.