Determining grain size distribution of the material composing landslide dams in the Northern Apennines: sampling and processing methods

Landslide dams are a particular kind of natural dam which forms when an earth or rock mass reaches a river channel causing its complete or partial blockage. Since 1995, the Earth Sciences Department of the University of Firenze has been running a project involving the inventory of landslide dams in the Northern Apennines, which currently contains more than 60 case studies. It is known that the geotechnical characterisation of landslide dams is closely related to the grain size distribution of the materials composing the dams. In particular, grain size exerts a control on dam stability, influencing the strength and the permeability of the dam material. The main purpose of this work is to contribute to grain size characterisation of landslide dams by giving an outline of the possible methods that can be used for grain size assessment and showing the results obtained from the case studies selected in the Northern Apennines. Indeed, despite its importance, there are few quantitative studies directed at determining grain size distribution of debris dams, mainly because customary bulk sieve analyses are impractical when dealing with materials often ranging in size from blocks tens of cubic metres in size, to microscopic particles. This paper presents results derived from the application of two main sampling methods on 42 landslide dams, (a) volumetric analysis and (b) grid by number analysis (here presented in two variants). These techniques allow us to obtain very similar grain size distributions for the coarser part of the debris material, while some differences persist for the finer fraction, The paper would also like to highlight the importance of getting accurate grain size data, with particular reference to the determination of the hydrograph derived from a landslide dam failure. Many deterministic models attempt to approach this problem, by taking into account only the median percentile (D-50) of the cumulate grain size distribution. However, in most of the selected cases, a marked bimodal distribution was found in the frequency diagrams, so the fiftieth percentile (D;()) is often one of the less representative grain size classes. It should thus be taken carefully into account that this choice could be misleading, as it is not necessarily justified by the physical evidence. (C) 2002 Elsevier Science B.V. All rights reserved.