Two years of humidity drift, measured against a salt test
2 April 2026
Two saturated salt solutions in sealed jars, left overnight at a stable indoor temperature: magnesium chloride at 33 % RH and sodium chloride at 75 % RH, both at 20 °C. The sensor that had spent two winters on the roof was tested next to one that had never left the drawer.
| Reference | Roof sensor | Drawer sensor |
|---|---|---|
| 33 % RH | 36.4 % | 33.5 % |
| 75 % RH | 79.1 % | 75.6 % |
So roughly +3.5 points at the low end and +4 at the high end — a nearly constant offset rather than a change of slope. That is consistent with contamination of the polymer layer rather than damage, and it is recoverable: after eight hours at 60 °C in a dry oven the roof sensor came back to within 1.2 points at both references.
Consequence for the archive
Humidity before the bake is not corrected retroactively in the stored data. A drift this gradual cannot be undone honestly with a single constant, and the raw record is more useful than a smoothed one. The reading date of each calibration is kept alongside the series instead.
The bake is now scheduled every autumn, before the first frost.