Articles | Volume 15, issue 9
Research article
22 Sep 2015
Research article |  | 22 Sep 2015

An approach to reduce mapping errors in the production of landslide inventory maps

M. Santangelo, I. Marchesini, F. Bucci, M. Cardinali, F. Fiorucci, and F. Guzzetti

Abstract. Landslide inventory maps (LIMs) show where landslides have occurred in an area, and provide information useful to different types of landslide studies, including susceptibility and hazard modelling and validation, risk assessment, erosion analyses, and to evaluate relationships between landslides and geological settings. Despite recent technological advancements, visual interpretation of aerial photographs (API) remains the most common method to prepare LIMs. In this work, we present a new semi-automatic procedure that makes use of GIS technology for the digitization of landslide data obtained through API. To test the procedure, and to compare it to a consolidated landslide mapping method, we prepared two LIMs starting from the same set of landslide API data, which were digitized (a) manually adopting a consolidated visual transfer method, and (b) adopting our new semi-automatic procedure. Results indicate that the new semi-automatic procedure (a) increases the interpreter's overall efficiency by a factor of 2, (b) reduces significantly the subjectivity introduced by the visual (manual) transfer of the landslide information to the digital database, resulting in more accurate LIMs. With the new procedure, the landslide positional error decreases with increasing landslide size, following a power-law. We expect that our work will help adopt standards for transferring landslide information from the aerial photographs to a digital landslide map, contributing to the production of accurate landslide maps.

Short summary
In this work, we present a new semi-automatic procedure to prepare landslide inventory maps that uses GIS applications and tools for the digitization of photo-interpreted data. Results show that the new semi-automatic procedure proves more efficient for the production of landslide inventories and results in the production of more accurate maps, compared to the manual procedure. The presented work has potential consequences for multiple applications of landslide studies.
Final-revised paper