Abstract: Air temperature is a basic element of conventional surface meteorological observations, and its method of observation and data errors are directly related to our understanding of atmospheric processes and forecast accuracy. Carrying out comparative analyses of different temperature observing systems and ensuring the accuracy and comparability of the observed data is important for the study of atmospheric science and the forecasting of weather and climate. Based on a side-by-side field comparison of a natural ventilation Louvred screen Temperature System (LTS) and an Aspirated shield Temperature System (ATS), from September 2009 to August 2010, the differences between the data obtained from them were analyzed. In addition, the relationship between the system biases and temperature, and the relationship between the radiation biases and solar radiation and wind speed, were discussed. Subsequently, a model to modify the system biases and radiation biases was proposed and, finally, the effects examined. The results showed that the LTS possessed warm biases of 0.19℃ and 0.29℃ at nighttime and during daytime, respectively; and the ambient-temperature-dependent systematic biases were a linear function of temperature, which increased by about 0.006℃ when temperature increased by 1℃. The ambient-solar-radiation- and ambient-wind-speed-dependent radiation biases were the result of the mutual coupling effect between solar radiation and environmental wind speed. With the warming effect, solar radiation was a good approximate parabolic function of the radiation biases; and with the cooling effect, the wind speed was a good approximate negative exponential function of the biases. After modification, both the biases at night and during the day dropped to 0.0℃. The consistency rate increased up to 92.3% and 96.0%, and the samples whose biases were between -0.2℃ and 0.2℃ increased from 64.5% and 45.3% to 83.7% and 80.6% at night and during the day, respectively.