Defining an optimal approach for landslide signal extraction from tree-ring anatomical parameter series: Extending the application capabilities of dendrogeomorphology

Abstract

Knowledge of past landslide movements is one of the parameters for defining landslide hazard or trigger analysis. Dendrogeomorphology, as one of the most accurate approaches to date landslides, is commonly based on macroscopic analysis of growth disturbances in tree-ring series of disturbed trees. However, the use of anatomical indices for landslide detection has not been properly tested, despite could be more sensitive than macroscopic approaches. This is because the lack of a methodology for extracting anatomical-based landslide signal. Thus, the aim of this study was to define an optimal procedure that will allow a wider use of anatomical analyses in landslide research. For this purpose, 41 disturbed beech (Fagus sylvatica L.) individuals growing on a monitored landslide with a known date of reactivation (May 2010) were sampled. A total of five procedures were designed and tested, based on defining a threshold for signal detection by calculating the mean value of the vessel lumen area (VLA) or its annual percentage change and subtracting the stdev. Of all the procedures tested, the one in which the threshold calculation was based solely on VLA values from the disturbed tree proved to be the most effective. This procedure correctly identified the known landslide year while introducing the least noise (4.0%) into the resulting chronology. A general potential threshold is a natural age-dependent increase in VLA values, however, due to the relatively young age of the trees analyzed, this effect did not show up. Thus, the proposed procedure can only be recommended for younger trees. This study has thus provided a new approach that will allow a wider and more efficient use of anatomical analyses in dendrogeomorphological research on landslides.