Hi there,
So I see this and think two things, Even fast or slow, Not enough precision to get anything useful in terms of trends and coralatible data IMO. You need more digits and perhaps some form of temperature compensation. or Just more Sensors. 
I would use a heater probe and do some Zero and Span clibrations then do some Multi-point calibrations and see what the curve looks like. Then try it in the in-situ environment for comparison to the base-line stuff you collect. (this is much like UHP gas analyzing for FABS)
my .02
HTH
GL 
PJ 