Use of THref data for QC/QA of RS92 Measurements and RH Corrections
RS92 Quality Assessment Example
RS92 Calibration Correction Example
The stable environment in the ventilated THref allows cases of a "bad sensor" like this to be identified. The two independent RH sensors (alternately heated) do not agree, leading to jumps in RH when the heated sensor becomes the measurement sensor. This behavior can be detected during the ventilation period by assessing the jump in RH when the heating cycle changes (U1-U2=-1.53% RH here). Much of the detailed RH structure during the sounding (black) is caused by a faulty sensor rather than true atmospheric structure, introducing additional uncertainty of an odd character. Also, there is only a 50-50 chance that the more accurate sensor will be the measurement sensor after the heating cycle ceases (at t=2300 s here). Such cases may not be suitable for case studies or statistical analyses or other uses.
This figure shows the same THref-RS92 comparison as Fig. 3, except that a correction for mean calibration bias [Miloshevich et al., 2009] has been applied to the prelaunch RS92 data (red). This reduces the RS92 RH measurements by the correction amount when t<0 (blue and green U1 and U2 measurements were not reduced). The result is that the RS92 RH bias relative to THref was reduced from +1.56% RH to -0.20% RH. A dataset of corrected RH measurements, as would result from correction of the dataset shown in Fig. 4, allows evaluation of a calibration correction and its dependences, and this can be tracked over time if sufficient data are available. Changes in the RS92 mean calibration bias (unlike time-lag and solar radiation errors) tend to occur at least annually when Vaisala re-calibrates references and/or implements sensor hardware or software changes that affect the RH measurements.