Global Positioning System (GPS) measurements through a field of km-size atmospheric boundary layer (ABL) inhomogeneities with a 10-ppm index of refraction excess have been simulated and inverted. Biases of up to 1–2 cm in height, 1–5 mm in horizontal, and ∼5 mm in zenith tropospheric delay (ZTD) are found, in either static or dynamic atmospheres, using 24-h solutions and estimating ZTD parameters. For 1-h sessions the scatter can increase by a factor of up to 5. These biases are attributed to the inadequacy of standard mapping functions. The use of numerical weather prediction (NWP) models and additional sounding techniques is discussed as a means of improving mapping functions. Raman lidars are thought to offer the highest potential for this purpose and for external calibration of both hydrostatic and wet path delay.