Quantifying wood decomposition by insects and fungi using computed tomography scanning and machine learning

0562731 - MBÚ 2023 RIV DE eng J - Článek v odborném periodiku
Seibold, S. - Mueller, J. - Allner, S. - Willner, M. - Baldrian, Petr - Ulyshen, M. D. - Brandl, R. - Baessler, C. - Hagge, J. - Mitesser, O.
Quantifying wood decomposition by insects and fungi using computed tomography scanning and machine learning.
Scientific Reports. Roč. 12, č. 1 (2022), č. článku 16150. ISSN 2045-2322. E-ISSN 2045-2322
Institucionální podpora: RVO:61388971
Klíčová slova: enzyme-activities * decay-rates * temperature * forest * larvae * coleoptera * climate
Obor OECD: Microbiology
Impakt faktor: 4.6, rok: 2022
Způsob publikování: Open access
https://www.nature.com/articles/s41598-022-20377-3

Wood decomposition is a central process contributing to global carbon and nutrient cycling. Quantifying the role of the major biotic agents of wood decomposition, i.e. insects and fungi, is thus important for a better understanding of this process. Methods to quantify wood decomposition, such as dry mass loss, suffer from several shortcomings, such as destructive sampling or subsampling. We developed and tested a new approach based on computed tomography (CT) scanning and semi-automatic image analysis of logs from a field experiment with manipulated beetle communities. We quantified the volume of beetle tunnels in wood and bark and the relative wood volume showing signs of fungal decay and compared both measures to classic approaches. The volume of beetle tunnels was correlated with dry mass loss and clearly reflected the differences between beetle functional groups. Fungal decay was identified with high accuracy and strongly correlated with ergosterol content. Our data show that this is a powerful approach to quantify wood decomposition by insects and fungi. In contrast to other methods, it is non-destructive, covers entire deadwood objects and provides spatially explicit information opening a wide range of research options. For the development of general models, we urge researchers to publish training data.
Trvalý link: https://hdl.handle.net/11104/0335221