Multipath error is of generally less consequence in GPS geodesy than is often supposed. The phase observable is subject to relatively small effects, revealed by typical postfit root-mean-square (RMS) residuals of <5 mm for dual frequency phase at permanent sites [e.g., Yunck et al, 1994], though peaks can reach 3 cm or more in hostile settings. The regular oscillation of the multipath signature together with long integration times can reduce the error to insignificance. Genrich and Bock [1992] note that an occupation time of 30 min is usually adequate for sub-millimeter multipath error and that 10 min can suffice for 1 mm performance even in difficult environments. Most vulnerable are rapid static and kinematic applications that require short-interval solutions in reflective settings.
Attempts to reduce the instantaneous multipath error often center on suppressing reflections at the site. A few groups are attempting to devise signal processing techniques to minimize sensitivity to reflections that reach the receiver. Meehan and Young [1992] describe a technique used in the SNR-8000 that exploits the differing effects in the early and late correlator channels to partly cancel multipath, with reported reductions of a factor of 2 to 4. An adaptive technique reported by van Nee and Siereveld [1993] shows promise but has yet to be demonstrated operationally. Another approach is to remove multipath during data analysis. Bishop et al [1994] report success exploiting the repeating daily multipath signature to form calibration templates. Wu et al [1993] noted that antennas having different spatial orientations (inevitable on long baselines) will measure different relative phases at the same transmit time (i.e., when sampling the same wavefront) that can persist at the level of 1 cm or more after double differencing. They provide a model of this ``phase windup'' which eliminates the anomaly and modestly improves residuals and solution stability. Young et al [1994] investigated phase variations in the GPS transmit antenna patterns and found that they are apparently inconsequential in precise geodesy.