Soil Loss Estimation Under Different Soil Management Using a Multispectral UAV

The objective of the work was to evaluate the potential of a multispectral drone in monitoring soil loss from agricultural plots subjected to different types of soil management. The study was conducted in spring 2024 on six plots (22 meters long, 4 to 8 meters wide and with a 16% slope) at the University of Perugia's SERLAB site. Two of the six plots were under conventional tillage (CTR); two with cover crops terminated with tillage (CCTx), and two with cover crops terminated with roller crimping (CCM). SERLAB is equipped to measure rainfall and soil loss from individual plots. Two UAV surveys were carried out: the first in April (after sunflower sowing) and the second in May after an erosive event. The images acquired by the UAV were processed using Agisoft Metashape Professional software version 1.7.5, employing the Structure from Motion (SfM) photogrammetric methodology. Point clouds were processed using Cloud Compare software (version 2.13) to produce raster-format Digital Surface Models (DSMs). An attempt to reduce the disturbance of the vegetation on the evaluation of the soil microtopographic variations was done by applying an NDVI-based filter to the point clouds and then obtaining corresponding Digital Terrain Models (DTMs). The soil loss was evaluated using the Digital model of the Differences (DoDs) between the two surveys. The comparison between the measured and estimated soil loss data proved satisfactory only for one of the two CTR plots, using the DODS of the DSMs. In the other plots, significant overestimates of soil loss were always obtained, attributable to the interferences generated by the vegetation. Generally, the vegetation filter based on the NDVI threshold did not yield the desired results. This may depend on the fact that in the period considered, part of the vegetation (especially in CCM plots) is present as a residue and, therefore, is not effectively distinguishable with an NDVI filter.