Systematic differences in aircraft and radiosonde temperatures

Abstract

Automated aircraft data are very important as input to numerical weather prediction (NWP) models because of their accuracy, large quantity, and extensive and different data coverage compared to radiosonde data. On average, aircraft mean temperature observation increments [MTOI; defined here as the observations minus the corresponding 6-h forecast (background)] are more positive (warmer) than radiosondes, especially around jet level. Temperatures from different model types of aircraft exhibit a large variance in MTOI that vary with both pressure and the phase of flight (POF), confirmed by collocation studies. This paper compares temperatures of aircraft and radiosondes by collocation and MTOI differences, along with discussing the pros and cons of each method, with neither providing an absolute truth. Arguments are presented for estimating bias corrections of aircraft temperatures before input into NWP models based on the difference of their MTOI and that of radiosondes, which tends to cancel systematic errors in the background while using the radiosondes as truth. These corrections are just estimates because radiosonde temperatures have uncertainty and the NCEP background has systematic errors, in particular an MTOI of almost 2°C at the tropopause that is attributable in part to vertical interpolation errors, which can be reduced by increasing model vertical resolution. The estimated temperature bias corrections are predominantly negative, of the order of 0.5°-1.0°C, with relatively small monthly changes, and often have vertically deep amplitudes. This study raises important issues pertaining to the NWP, aviation, and climate communities. Further metadata from the aviation community, field experiments comparing temperature measurements, and input from other NWP centers are recommended for refining bias corrections.