Comparison of aircraft and radiosonde temperature biases at NCEP

Routine monitoring of temperature biases versus the NCEP guess show that globally averaged radiosonde temperatures are a few tenths of a degree colder than the guess in the pressure range 300 to 200 hPa. On the contrary, aircraft temperature data from ACARS, AMDAR and AIREPS in the same range of pressure show a warm bias of similar magnitude. These bias differences are fairly constant for each month of the year and all 4 synoptic cycles of the day and are large enough to require investigation and possible corrective measures. Collocation statistics of ACARS versus radiosonde data confirm this bias difference without requiring a guess. Our detailed analysis shows that the NCEP radiation correction is making the radiosonde cold bias larger and also, for certain types of sondes, makes the diurnal temperature bias larger. Diagnostics are shown confirming the work of (Redder et al. 2003) that the incorrect radiation correction applied at US Vaisala RS80 sonde sites makes the diurnal differences in temperatures excessive. Analysis of aircraft temperatures shows that the temperature bias distribution versus the guess is rather asymmetrical and non-Gaussian. Many more aircraft have large warm biases than cold. Communication with airline meteorologists and engineers indicates that dirt, debris, insects etc. in the aircraft temperature sensing tubes all result in warm biases. Since this is a physical explanation for excessive warm biases being spurious, one possible solution to reduce the average warm bias for aircraft data that pass quality control would be to reduce the bias limit for what aircraft units are on our reject-list due to an excessive long-term temperature bias.

Redder, C., J. Luers, and, R. Eskridge, 2003: Unexplained Discontinuity in the U.S. Radiosonde Temperature Data. Part II: Stratosphere. J. Atmos. and Oceanic Technol., 21, 1133-1144.