Abstract
Landslides of the slide-type movement represent common damaging phenomena in
the Italian province of South Tyrol. Up to January 2019, the landslide inventory of the
province lists 1928 accurately mapped landslides that required intervention by e.g. the
local road service or the provincial geological survey. Thus, this landslide data set
mainly includes events that caused damage. The aim of this contribution was to
investigate and critically interpret statistical associations between the inventoried slidetype
movements and a variety of spatial environmental variables. The assessment of
conditional frequencies and the discriminatory power of single variables revealed
conditions that are typically present at landslide mapping locations, e.g. topography,
land cover, rock types, and proximity to infrastructure. A critical interpretation of the
statistical results highlighted the need to consider the landslide data origin (i.e.
background information) in order to avoid misleading statements and wrong
inferences. The findings of the here presented work show that the availability of
detailed landslide information does not always ensure that valid process-related
conclusions can be drawn from subsequent statistical analyses (e.g. identification of
important landslide controls). Despite considerable methodical advancements in the
field of statistical data analysis and machine learning, we conclude that the principle
‘correlation does not necessarily imply (geomorphic) causation’ remains of particular
relevance when exploiting available landslide information.