
RS92 RH and T Accuracy Relative to THref From a Dataset of ComparisonsThis dataset of RS92THref comparisons is from the first THref deployment during the MOHAVE 2009 Raman lidar and satellite validation campaign at the NASA/JPL Table Mountain Facility. The RS92 RH and T bias relative to THref is shown as a function of the measured variable for 41 comparisons. The mean and standard deviation of the bias is given by the numbers and "error bars" at left (blue for nighttime and red for daytime soundings). The left and middle panels show the RH bias as a function of RH in two different ways, and the right panel shows the RS92 T bias as a function of T.
A dataset of RS92THref comparisons gives the mean RS92 calibration bias and its dependence(s) at the surface over the range of RH and T conditions measured, in addition to the individual sensor biases. With sufficient statistics, this information can be used to derive a correction for the mean calibration bias at the surface (see Miloshevich et al. [2009], section 2.4 and Figs. 7 and 13). This information is also useful for characterizing RS92 sondetosonde variability within and between calibration batches, as well as tracking of manufacturer changes in the sensor hardware or the calibration function over time, which is important for longterm monitoring applications. There is also some promise for the idea of correcting RS92 T (and maybe P) measurements for individual sensor bias (as opposed to mean calibration bias) by subtracting the individual sensor bias from the measurements (normalizing to THref), which would eliminate sensortosensor variability and guarantee absolute accuracy within ±0.1°C, but only if it can be shown that the bias is independent of T over the full T range of a sounding (requires further investigation). 