4D monitoring of mountain areas using the UAV-PPK workflow

Abstract

The mapping and three-dimensional modeling of mountain areas and the study of land cover change dynamics are current tasks in preserving and maintaining protected natural parks and forests. In this context, recent developments in digital photogrammetry using the SfM-MVS method to process captured imagery and the development of unmanned aerial systems (UAS) allow for reducing the costs, time, and the use of human resources and obtaining and repeatable 3D topographic data for moun-tainous regions. We will call this acquired 3D high-resolution topographic data (HRTD) 4D data in the context of an additional temporal component. The main objective of this study is to evaluate the applicability of PPK (Post-Processing Kinematic) direct georeferencing of images captured by UAVs and processed through the SfM-MVS method to obtain HRTDs for 4D land cover analysis. We analyze a 3D HRTD with an acquisition interval of two years for a mountain test area in Plana Mountain near Sofia. The test area has a diverse vegetation cover, including coniferous forest, grassland, hay meadows, shrubs, and single deciduous trees. We conducted multiple surveys of the test area with a budget PPK-UAV configuration (DJI Phantom 4 Pro with a single-frequency PPK-GNSS kit installed) from March 2020 to October 2022. Two autumn surveys from September 2020 and October 2022 were se-lected, which possess the most-good performance on numerical data accuracy. We performed 3D data analysis on 1) Assessment of the accuracy of PPK-SfM-MVS photogrammetry generated topographic data (3D clouds and DSM); 2) Investigation of the errors in the individual specific surfaces (for the individual isolated sections) using the M3C2 tool for comparing and evaluating dense point clouds; 3) Determining land cover changes in the demarcated areas using a surface of differences (DoD). Accuracy analysis showed that the PPK solution provides comparable accuracy (about RMSE3D = 0.067 m for the 2020 data, georeferencing (PPK+1GCP) and RMSE3D about 0.13 m for the 2022 data, georeferencing (PPK only)) like the GCP method. The multi-temporal topographic reconstructions based on UAV- PPK-SfM allowed us to quantify and qualitatively determine the land cover changes that occurred. The UAV-PPK-SfM workflow in the context of 4D land surface monitoring and the results suggested that even low-cost UAV-PPK systems can provide data suitable for measuring geomorphic change at the scale of the acquired data. The multi-temporal topographic reconstructions based on UAV- PPK-SfM allowed us to quantify and qualitatively determine the land cover changes that occurred. The UAV-PPK-SfM workflow in the context of 4D land surface monitoring and the results suggested that even low-cost UAV-PPK systems can provide data suitable for measuring geomorphic change at the scale of the acquired data.