Accuracy of gully morphological parameters extracted by UAV photogrammetry in the Loess Plateau

Unmanned aerial vehicle (UAV) and ground Lidar scanning are 2 new methods used for soil erosion investigation in recent years．However, the accuracy of UAV-generated topography in areas with complex terrain is unclear．To explore applicability of UAV photogrametry in gully erosion monitoring, gully morphological parameters extracted from UAV images by a fixed-wing drone at a flight height of 122 m were compared to ground Lidar scanning data in a small catchment on the hilly and gully Loess Plateau．Percentage errors in 2D gully parameters including area, perimeter and length extracted from UAV image data were found to be <10%．UAV image-extracted 3D gully parameters (ross sectional area, average gully depth and gully volume) showed errors of <20%, with lower errors for small gullies with a gentle gully sidewall and sparse vegetation, higher errors for larger gullies with steep gully sidewall and high vegetation coverage．Errors in gully parameters measured by UAV photogrametry were mainly due to the fact that fixed-wing drones collect data vertically, leading to the loss of sidewall details thus reducing accuracy of parameter extraction．Vegetation showed certain influence on measurement accuracy, due to lack of effective vegetation filtering．Compared with ground Lidar scanning, accuracy of geomorphological parameters from UAV photogrametry was found to be slightly lower, implying limitations in short-term monitoring of gully erosion．