The sensor's specifications must provide metrological requirements for the entire test period. Or you should create a scheme that would provide the required metrological characteristics for the entire test time. In this case there should be no problems.

Michael,

Sensors exist that show little drift over such timescales.

I wouldn't bother with NDIR as a modality, for example, but simple catalytic bead sensors.

You might have some oddball temp/RH calibrations to do, and in general you get what you pay for (Draeger makes some nice kit) but in principle they should work.

I recommend to use NDIR sensor.

Advantages

- long lifetime

- stability

- many types and producers

For example: http://www.alphasense.com/WEB1213/wp-content/uploads/2015/05/IRM-AT.pdf