Bad data was used in the analysis from 860101 00Z to 860102 12Z inclusive due to an error in the earth location of the soundings at NOAA/NESDIS, Washington. No satellite soundings were used in the analysis until 4 January 1986, when the problem was resolved.