The evaluation of unmanned aerial system-based photogrammetry and terrestrial laser scanning to generate DEMs of agricultural watersheds

Abstract

Agricultural watersheds tend to be places of intensive farming activities that permanently modify their microtopography. The surface characteristics of the soil vary depending on the crops that are cultivated in these areas. Agricultural soil microtopography plays an important role in the quantification of runoff and sediment transport because the presence of crops, crop residues, furrows and ridges may impact the direction of water flow. To better assess such phenomena, 3-D reconstructions of high-resolution agricultural watershed topography are essential. Fine-resolution topographic data collection technologies can be used to discern highly detailed elevation variability in these areas. Knowledge of the strengths and weaknesses of existing technologies used for data collection on agricultural watersheds may be helpful in choosing an appropriate technology. This study assesses the suitability of terrestrial laser scanning (TLS) and unmanned aerial system (UAS) photogrammetry for collecting the fine-resolution topographic data required to generate accurate, high-resolution digital elevation models (DEMs) in a small watershed area (12ha). Because of farming activity, 14 TLS scans (≈25pointsm−2) were collected without using high-definition surveying (HDS) targets, which are generally used to mesh adjacent scans. To evaluate the accuracy of the DEMs created from the TLS scan data, 1098 ground control points (GCPs) were surveyed using a real time kinematic global positioning system (RTK-GPS). Linear regressions were then applied to each DEM to remove vertical errors from the TLS point elevations, errors caused by the non-perpendicularity of the scanner's vertical axis to the local horizontal plane, and errors correlated with the distance to the scanner's position. The scans were then meshed to generate a DEMTLS with a 1×1m spatial resolution. The Agisoft PhotoScan and MicMac software packages were used to process the aerial photographs and generate a DEMPSC (Agisoft PhotoScan) and DEMMCM (MicMac), respectively, with spatial resolutions of 1×1m. Comparing the DEMs with the 1098 GCPs showed that the DEMTLS was the most accurate data product, with a root mean square error (RMSE) of 4.5cm, followed by the DEMMCM and the DEMPSC, which had RMSE values of 9.0 and 13.9cm, respectively. The DEMPSC had absolute errors along the border of the study area that ranged from 15.0 to 52.0cm, indicating the presence of systematic errors. Although the derived DEMMCM was accurate, an error analysis along a transect showed that the errors in the DEMMCM data tended to increase in areas of lower elevation. Compared with TLS, UAS is a promising tool for data collection because of its flexibility and low operational cost. However, improvements are needed in the photogrammetric processing of the aerial photographs to remove non-linear distortions.