Digging up into windstorms aftermath: understanding the effect of harvesting systems on salvage logging wood residues spatial distribution

Extreme disturbance events, such as climate change-driven ones, have increased their frequency, upsetting the ordinary management of forests, and impacting large areas with severe damage. As a consequence, when productive forests are hit, salvage logging operations represent the common way to recover part of the economic loose. However, salvage logging can lead to negative impacts in terms of soil erosion as well in terms of variation of soil carbon stock and nutrients. Commonly, in the European Alps, salvage logging operations in largely damaged forest areas can be referred generally to as two harvesting systems: i) Cut-to-Length (CTL) and ii) Full-Tree (FT) extraction systems. The application of the two harvesting systems can have a different effect on the type and quantity of logging residues and deadwood left on the forest ground, which in the short-medium term it can be reflected in terms of quantity and distribution of organic carbon and nutrients in the soil. To evaluate and gather more detailed information on the effects of forest operations, a valuable option is to rely on the use of precision forestry approaches, such as the use of remote sensing (RS) and Artificial Intelligence applications, for example, machine learning (ML). In the realm of forest operations, ML techniques and algorithms are the most used and can be fed with data directly extracted from the machines operating live on sites, and also with data retrieved through RS. Drone-borne data, for example, is now becoming the most used for its large potential and applicability, providing both large coverage and a high level of detail oversampled areas. The aim of this study is thus comparing two salvage logging areas and find any difference in terms of logging residue type, quantity, and spatial distribution according to the used harvesting system. Drone flights over two logging areas in the northeastern Alps to retrieve logging residues data were performed in 2022, the sites were windthrown in 2018 and harvested in 2021. A random forest model was built using RGB bands derived from the drone images, textural variables, and information from the surface model to classify elements in a clear-felled site. After the classification and noise removing operations, residues mass per hectare and distribution were estimated. Preliminary results will show the strengths and weaknesses of the method adopted in assessing the type of residues and their spatial distribution. Moreover, this application will highlight the different impacts of the two systems adopted in salvage logging operations with respect to residue type and quantity left on site.