We analyze the effect of small-scale refractivity inhomogeneities, associated with the water vapor field in the lower troposphere, on GPS station height determination. Numerical simulations of GPS signals have been performed in several model atmospheres and inverted using a continuous satellite distribution approach and Bernese GPS software. Results show cm-level errors in station heights, when ZTD parameters are estimated with standard mapping functions. The errors are due to mismodeling of the refractivity distribution in the troposphere. A means to calibrate properly humidity inhomogeneities is to sense the wet path delay with a water vapor radiometer (WVR) or a lidar. Numerical simulations based on radiosoundings show that wet path delay retrieved with WVRs can exhibit up to ∼10 mm biases, due to temperature inversions and high humidity contents. Lidars show higher accuracy, even when only the water vapor concentration is measured (neglecting the effect of temperature inversions). In order to achieve the highest accuracy from GPS, we propose to combine a Raman lidar data for external wet path delay calibration and estimate the remaining path delays during GPS data analysis. Since the latter is quasi-homogeneous, once the highly inhomogeneous humidity component is removed, this should indeed improve the positioning accuracy. Numerical weather prediction (NWP) models might also be used to replace standard mapping functions, especially to estimate the hydrostatic path delay component.