Abstract: Based on the daily temperature data from 75 radiosonde stations in the eastern China at 08:00 and 20:00 during 1958 to 2005 and the corresponding metadata, homogeneity of temperature at different levels was tested and adjusted using  methods of a temperature difference between daytime and nighttime and a two-phase linear regression. The annual trends of the observed and revised temperature at 49 stations were analyzed by selecting data using 30 % maximum missing rate as a standard. The results indicate that inhomogeneity of temperature is caused by changes of radiosonde instrument and radiation correction method in upper level over the eastern China. The credible breakpoints for inhomogeneity occur in 1966 and 2000, especially in 1996. The most obvious inhomogeneity appears at 100 hPa. Compared with the original temperature series, warming trend of temperature below 500 hPa weakens from 1958 to 2005, so is its decreasing trend from 200 hPa to 50 hPa and the ratio is 0.04-0.08 ℃ every decade. Troposphere temperature increases and lower stratosphere temperature decreases significantly since 1979. Therefore, temperature discrepancy of adjusted and original time series during 1958 to 2005 is small compared with that during 1958 to 2005.

Key words: Upper air temperature, Inhomogeneous adjustment, Day-night temperature difference, Two-phase linear regression