Precipitable water vapor (PWV) is an essential parameter in numerical weather prediction and climate research. Existing global empirical PWV models rely on a single coefficient for vertical adjustment and lack geographical differentiation. Therefore, this study developed the global PWV vertical adjustment model (GPWV-H) by considering the time-varying lapse rate using the fifth-generation European Centre for Medium-Range Weather Forecasts Atmospheric Reanalysis (ERA5) from 2012 to 2017. The performance of the GPWV-H model in vertical adjustment is evaluated using multi-source PWV data and compared with the conventional empirical model (EPWV-H). The numerical results are as follows: (1) The bias and root mean square (RMS) of the GPWV-H model are − 0.10/ − 0.35 mm and 1.43/1.07 mm, respectively, when ERA5 and radiosonde PWV profiles were used as reference which are 9.3 and 5.9% (in RMS) lower than EPWV-H model; (2) The GPWV-H model improved by 15.1–17.1 and 0.8–1.6% compared to the non-adjustment and the EPWV-H model, respectively, when interpolating Second Modern-Era Retrospective Analysis for Research and Applications (MERRA-2) with various grid resolutions to radiosonde stations. These results indicate that the GPWV-H model outperforms the EPWV-H model regarding global PWV interpolation accuracy and stability and has a promising application tendency in global real-time and high-precision water vapor monitoring.