Abstract: The weighted mean tropospheric temperature is a critical parameter in the conversion of wet zenith delay to precipitable water vapor in GPS meteorology.Four existing methods, namely, approximate close form, constant method, Bevis method and numerical integration, were developed to perform this conversion.However, the approximate close form is very difficult to realize due to their three assumptions; the constant method induces a large error in the conversion.The Bevis method has a systematic error in the calculation of weighted mean temperature and the numerical integration cannot provide the weighted mean temperature in real-time from the radiosonde data because of the restriction of maintenance costs.The conversion parameter from wet zenith delay to precipitable water vapor is a function of weighted mean tropospheric temperature.We find that the admissible error of weighted mean temperature should be smaller than 3.4K if the conversion accuracy is better than 1mm.The weighted mean temperature is strongly correlated to the surface meteorological measurement based on the correlation analysis of 13 months surface and upper-air meteorological data in Hong Kong.An empirical formula, which is a function of surface temperature, is established using sequential regression analysis method based on these surface and upper-air meteorological measurements.Another two months data are used to demonstrate the validation of this new empirical formula.This real-time formula satisfies the requirement of both real-time and accuracy in GPS meteorology, and it is able to eliminate the systematic error caused by the Bevis method in the Hong Kong region.These results also exhibit that the empirical formula for weighted mean tropospheric temperature should be determined using regional meteorological measurements in the GPS meteorology application.