Open Derek-Hageman-NOAA opened 4 years ago
I am all about doing a better measurement and providing data that has been processed consistently. @gmd-jw ... do you have an opinion about this. Did only SPO use the specialized four parameter fit. Barrow does get cold to!?!
All baseline stations used the 4-parameter fit from Ells. I found an old plot comparing the effect of switching to the Steinhart-Hart fit on PIR data. At -80C there is only a 0.02W/m^2 difference. There is a similar difference at +40C. I say that the difference is insignificant compared to other uncertainties (ie. 0.1C interchangeable thermistors).
I would guess that most people are not using Ells's equation, and I have seen the Steinhart-Hart constants published in a document from BSRN. Even though we used Ells's equation for many years, I would use either the actual table data Derek referred to, or the Steinhart-Hart fit, and document the change.
Currently all thermistors are converted using a simple Steinhart-Hart fit (3 parameter). This is known to be less than ideal at very low (or high) temperatures.
The older system used a specialized four parameter fit for SPO to account for the low tempertures. This is not currently implemented in the newer processing. I would like to switch all conversion to using an interpolated table lookup from the manufacturer specifications. That is, the manufacturer provides a table of resistances at 1 degree intervals and the system can interpolate between the two bounding resistances to get the final output temperature. As far as I know, this should exceed the accuracy of either fit. This is available to enabled right now relatively trivially.
However, changing this does represent a processing change, so I haven't currently done it. If desired instead, I could also implement the four parameter fit for SPO only (which would match the older processing).