Assessment of digital terrain models for multipath prediction at geodetic GNSS installations

This paper describes multipath modeling procedures and results for a permanent geodetic GNSS site. The site is located in an environment where multipath is predominantly caused by reflections from the surrounding terrain and it is the goal of this study to assess different terrain models for multipath prediction in this kind of environment. Two terrain models are created. One is derived from the National Elevation Dataset (NED) and is accurate to ∼10 m in the vertical, while the other is based on Light Distance and Range (LiDAR) data and is accurate to ∼20 cm. Experimental data logged at the site are used for model validation and the results are given in the form of overall error statistics and per satellite time history comparisons. For the LiDAR model, the elevation dependent error statistics match to within 11 cm or better and C/A code multipath time histories generally show good agreement in terms of amplitudes and frequencies. Several useful visualizations based on the modeled results are also presented. They show where multipath reflections occur in this environment and relate them to relative amplitude and geometric delay. The visualizations reveal that multipath predominantly occurs around hills and ridges within 500 m of the site and that most signals have relative amplitudes of less than 0.1 and delays of less than 30 m. While small amplitudes and short delays typically yield small tracking errors, the results for this site show that 1-2 meter level errors nonetheless occur. A comparison between narrow and double delta correlators shows that multipath performance improvements are not seen for low and high satellite elevations because these are regions where geometric delays are typically a few meters or less. Finally, time history comparisons for the experimental data, LiDAR model simulation, and the NED simulation show that only the high accuracy terrain model is able to provide consistent and accurate multipath predictions.