reachable accuracy with multi-point calibration

[soerenthomsen]

At the end of Section Calibration equation and firmware versions we write: During/between field deployments there are possibilities for end users to post-adjust the sensors either by a one-point air-saturation adjustment or by taking reference samples (e.g. water samples and Winkler titration) and/or using a well-calibrated sensor in parallel. If done correctly such an adjustment should result in an absolute accuracy of around 1 % for multipoint calibrated sensors (model: 4330, 4831, 5331 and 5730) and 3 % for two-point calibrated (model: 4835, 4531), see below for more information about factory calibrations.

@hahn-johannes raised the point that this might be "quite optimistic, when considering the whole Temperature-Phase-O2 space."

@tomhull agreed and said real user examples would be good to have

@hahn-johannes "we could take some numbers from Bittig et al. (2018), who did a lot on the evaluation of multipoint calibration, batch calibration and two-point calibration. Further, in Hahn et al. (2017) (not an optode calibration paper), in situ calibrations against CTD measurements during a CTD-cast should an rms of 3 µmol/kg.

For the presentation of calibration errors, it seems to me that they are given in units of µmol/kg rather than saturation. I suggest to use this."

[soerenthomsen]

It would be great if we could agree on the formulation during the community review.