Simplified calibration process and uncertainty assessment for sampling large numbers of single-use thermistors for upper-air temperature measurement

Abstract

Thermistors for upper-air temperature measurement are commonly used only once in radiosonde sounding. Considering that large numbers of thermistors are used for networks of measurements and that the thermistors are cheap, the calibration of individual thermistors at several calibration points is considered impractical. In this work, due to economic considerations, a method in which a subset of thermistors are selected for calibration was devised, and the uncertainty attributed to the sampling process was assessed. After having calibrated 32 thermistors at full calibration points between –70 °C and 40 °C, we demonstrated a simplified process and uncertainty assessment that involve sampling of only a few thermistors for the full calibration. From the average temperature deviation and the standard deviation s of for the five sampled thermistors expressed in terms of the temperature differences and their dispersion, the expanded interval (−5s, + 5s) can include more than 95% of the thermistors with a level of confidence of 95%. From a simple simulation, assuming that the characteristics of thermistors from the same batch are normally distributed, the coverage factor to be multiplied by the sample standard deviation can be calculated for different sample sizes, target fractions of thermistors to be covered, and levels of confidence. This technique can be extended to efficiently calibrate a large number of single-use sensors from the same batch, and it can be an economical solution for finding a balance between accuracy and efficiency in such cases.