Unexplained Discontinuity in the U.S. Radiosonde Temperature Data. Part II: Stratosphere

Abstract

The second part of this two-part article discusses the differences between observations taken at 0000 and 1200 UTC, particularly in the stratosphere, by the Vaisala RS80-57H radiosondes that are integrated within the National Weather Service's (NWS's) Micro-ART system. There are large maxima in the horizontal distributions of the monthly time means of the 0000/1200 UTC temperature and height differences over the central United States that are absent over Canada. These maxima are as large as 5 K and 150 m at 10 hPa. Data analysis shows that the 0000/1200 UTC differences are largely artificial, especially over the central United States. They originate in the postprocessing software at observing stations, thus confirming the findings in Part I. Special flight data from the NWS test facility at Sterling, Virginia, have been used to deduce the bias correction applied by Vaisala's postprocessing system. By analyzing the correction data, it has been shown that the inconsistencies with non-U.S. Vaisala RS80 data, as well as most of the large 0000/1200 UTC differences over the United States, can be accounted for by multiplying the reported time since radiosonde launch by a factor of 5/3, which is incorrectly applied by the Vaisala postprocessing software. After being presented with the findings in this paper, Vaisala further isolated the source of the inconsistencies to a software coding error in the radiation bias correction scheme. The error affects only the software installed at NWS stations.