Impact of mapping errors on the reliability of landslide hazard maps

TLDR

Mapping landslide deposits is intrinsically difficult and subjective, requiring significant effort to reduce inherent uncertainty. The study aims to evaluate how mapping errors affect landslide hazard predictions by building discriminant models using geological‑geomorphological predictors and landslide occurrences from each inventory map. Three independent inventory maps of the 300‑km² Staffora Basin were produced, and discriminant models were built using the same geological‑geomorphological predictors and landslide occurrences from each map. Comparisons revealed large positional discrepancies (55–65%) between maps, increasing to over 80% when all three overlapped, yet statistical modeling largely mitigated the impact of these errors, though data inaccuracies remain a major limitation to hazard map reliability. Abstract.

Abstract

Abstract. Identification and mapping of landslide deposits are an intrinsically difficult and subjective operation that requires a great effort to minimise the inherent uncertainty. For the Staffora Basin, which extends for almost 300 km2 in the northern Apennines, three landslide inventory maps were independently produced by three groups of geomorphologists. In comparing each map with the others, large positional discrepancies arise (in the range of 55–65%). When all three maps are overlain, the locational mismatch of landslide deposit polygons increases to over 80%. To assess the impact of these errors on predictive models of landslide hazard, for the study area discriminant models were built up from the same set of geological-geomorphological factors as predictors, and the occurrence of landslide deposits within each terrain-unit, derived from each inventory map, as dependent variable. The comparison of these models demonstrates that statistical modelling greatly minimises the impact of input data errors which remain, however, a major limitation on the reliability of landslide hazard maps.