*3.2. Monitoring of Morpho-Topographic Changes in the Test Areas Based on UAV-Derived Data*

The monitoring of the short-term evolution of the test areas was based on annual UAV surveys. To date, three UAV survey campaigns have been carried out—in July 2019, 2020, and 2021; a Phantom 3 Standard (Quadcopter) was used, which is a drone developed by Da-Jiang Innovations (DJI, Shenzhen, China). This drone was mounted with a stabilized camera that compensated for involuntary movements of the UAV due to wind, thus ensuring the correct orientation of the photos with respect to the ground. The coordinates for each photo were then registered by an internal GPS, allowing the drone to follow the points (waypoints) previously fixed with the GPS on the flight plan. An average flight altitude of 40 m was used for all surveys, which allowed for a ground resolution of 1.5 cm/px. Using the default camera of the Phantom 3 Standard, about 250 images with a longitudinal overlap (flight direction) of 85% and a flight strip overlap of 60% were obtained during each flight.

For all flights, six targets, 40 × 40 cm in size and easily visible from above, were placed to record the position of the Ground Control Points (GCPs) and consequently to orient the model in space. A GNSS receiver in static nRTK mode was used to acquire the GCPs.

To evaluate the variations in the plano-altimetric features of the test areas and the related volumetric changes, 3D models were generated. To obtain these models, the (2D) photos acquired with the drone were processed using the Structure from Motion (SfM) algorithm, which internally implements the photogrammetry and computer vision methods. A functional correlation between the 3D object points and the 2D image points via the collinearity condition is the core concept behind photogrammetric image data processing [35]. The examination of two photos and related orientation parameters allowed for the identification of the common points and the determination of the related 3D coordinates.

To obtain the final outputs (Dense Cloud Points, DEMs, and orthophotos), we proceeded as follows: (i) generation of the flight plan; (ii) field work consisting in the positioning of the targets, measurement of the position with the GNSS receiver, and acquisition of photos with the UAV; (iii) aerial data processing (as described by Snavely et al. (2008)) [36], error checking of the model against GCPs, and export of point clouds, DEMs, and orthophotos.

Minervino Amodio et al. (2022) [14] already highlighted that there is a strong correlation between the points surveyed on the beach with the GNSS and the UAV techniques. This correlation allowed us to directly use the DEMs 2019, 2020, and 2021 obtained from the UAV surveys for extracting the beach profiles needed for our investigation. Overall, 12 beach profiles, extending from the top of the dune ridge up to the water line (see below), were extracted.
