EXAMINING TERRESTRIAL LIDAR ACCURACY IN THE BADLANDS: A CASE STUDY AND GENERAL LESSONS
Compared to traditional topographic mapping tools, airborne and terrestrial lidar present unique challenges for assessing the accuracy of raw data products. A large number of internal and external factors that cannot be controlled and may be difficult to measure contribute to the error model associated with each lidar measurement. As a result, it is virtually impossible to make reliable a priori estimates of the uncertainty associated with each measurement. In addition, unlike traditional surveying and photogrametry, only a single measurement is made to each point so it is impossible to make a posteriori estimates of uncertainty based on network adjustments. Therefore, it is necessary to statistically evaluate the accuracy of lidar data using independent checkpoints of higher accuracy.
The error analysis was performed by measuring a small number of checkpoints (n≈20) with a reflectorless total station at several of the larger sites used in our study. Each site consisted of a several thousand square meter hillslope that was typically scanned from a single position 50-100 meters from all points on the slope. A more detailed analysis of a single 1700 m2 hillslope was also performed using 385 checkpoints in order to investigate the spatial distribution of errors.
Initial results from these analyses indicate that, at the 95 % confidence level, measured residuals are approximately two times the values estimated from the scanner manufacturer's specifications alone. More importantly, the spatial distribution of residuals is not isotropic. In addition, due to the limitations of line-of-sight measurements, deep rills that incise parts of the slopes were not completely represented. Additional systematic errors due to environmental conditions were also observed in some scans. Nonetheless, the ability to remotely map hillslopes without introducing physical disturbances made the parent erosion study possible.