4.3
The identification of observation errors in radiosonde mass data at the National Centers for Environmental Prediction
William G. Collins, NOAA/NWS/NCEP/EMC, Washington, DC
The term "observation errors" is used in the special sense of referring to the combination of measurement or other random errors and large non-random rough errors, present in the radiosonde data prior to processing at the observation station. These errors represent one of the kinds that is diagnosed by the National Centers for Environmental Prediction s complex quality control (cqc) program for radiosonde data, and they are the ones most difficult to properly identify.
The observation errors, as defined above, are modeled as a combination of a random and non-random component. The principles of cqc are used in the diagnosis of observation errors, with checks on the increment (observation-forecast), horizontal interpolation of increment, and vertical interpolation of increment giving the check residuals. Using statistics on global values of the increments for data not initially suspected of error, a combination of the error modeling and residual values gives a likelihood of observation error for each height or temperature datum.
Examples are shown to illustrate the method and its performance over a period of time is described. The limitations of the method are outlined.
Session 4, Radiosondes and Rawinsondes
Monday, 15 January 2001, 3:30 PM-5:30 PM
Previous paper Next paper