Keeping the record straight

Environmental measurements are not fixed. A reading published this morning may be revised this afternoon, replaced overnight, or removed entirely when a station is recalibrated or retired. This is normal practice and usually an improvement \u2014 the revised figure is often the better one. The problem is not revision. The problem is that the earlier figure usually disappears, and with it any way to answer a simple question: what did the record say on the day it mattered?

That question is the whole reason this platform exists. If someone had a bad day in March, the useful comparison is against what the air was measured to be in March, as reported at the time \u2014 not against a figure that has since been smoothed, re-based or replaced.

What we do instead

Every reading is written once, dated, and kept. When an upstream source later publishes a different value for the same place and time, we keep both and record when each was read. Neither is deleted. Where the two disagree materially, that disagreement is itself a fact about the record, and it is shown rather than resolved silently in favour of the newer number.

The same answer tomorrow, with a receipt.

And where there is nothing

The harder case is absence. Large areas have no nearby monitor at all. The standard response is to interpolate \u2014 to model a value from the nearest stations and present it as though it were measured. That produces a complete-looking map, and a misleading one.

We would rather say the gap exists. If the nearest reference monitor is twenty kilometres away, the honest output names that distance instead of filling the space with an estimate dressed as a reading. Where we do model something, it is labelled as modelled and never mixed into the measured record.

Why this is the first post

Because everything else here depends on it. Future posts will make claims about places, seasons and trends, and those claims are only worth anything if the underlying record is stable, dated and checkable. This post is the standard the rest will be held to \u2014 including by you. If something here is wrong, say so and it gets corrected in place, with the correction marked.