Some considerations about the simulation of the breach channel erosion on landslide dams

The analysis of the flood hazard related to the downstream areas of the blockage and generated by a possible failure of the landslide body is one of the most interestig aspects of the landslide dams study. The BREACH code (Fread,1984), simulating the collapse of earthen dams both man-made and naturally formed by a landslide, has been chosen in order to analize this case and it has been applied to a real landslide dam: the landslide of Valderchia, (central Italy), already subject of research (Cencetti et alii,1998). Amoung the geotechnical parameters, required as input from the model, the mean diameter, the D50 and the ratio D90/D30 are difficult to determinate in case of landslide dam bodies. Indeed, it is often impossible to determinate the proper grain size composition of the landslide material, using the routine methodologies of investigation (sampling via geognostic drillers) because it is not reckoned in the due way the presence of coarse materials as boulders can be. Better results, instead, can be obtained using tipical methods of gravel bed rivers sampling and analysis (Ermini & Rosati, 2002). The BREACH code simulates the break of a dam assuming that the size of the breach along the crest of the landslide dam is governed by the capacity of the flowing water (which erodes and carves the bed of the breach channel) to transport the eroded materials. The bed load transport formula used in BREACH (Meyer-Peter & Muller, modified by Smart (1984), is based on experiments performed in a flume with the grain size distribution ratio D90/D30 lesser than 10. Such a methodology probably makes this equation not much suitable to describe the sediment transport peculiar to a landslide body having a very low sorting. So, in accordance with the results of the experiments carried out by Bathurst et alii (1987) both on flume and mountain river beds, the Schoklitsch formula (Schoklitsch, 1962) has been implemented into the program, as an alternative to the Smart equation. The comparison between simulations, alternatively applying the equations of Smart or Schoklitsch, shows that the latter (Schoklitsch formula) seems to work better. However, because the landslide bodies often have a strongly bimodal grain-size frequency curve, , the percentile D50 (the tipical granulometric parameter requested by the bedload sediment transport formulas) can correspond to one of the grain-size classes which are less present in the reality. Moreover, in accordance with Wilcock's observations (Wilcock, 1992), in the case of strongly bimodal grain-size distributions, the finer fraction is set in motion by a lesser shear stress than the coarse fraction. This, in the case of a breach formation (for overtopping) along the top of a landslide dam, can determine the armouring of the breach bed, with a resulting stop of the erosion. In order to simulate this phenomenon, the BREACH program has been implemented with a new procedure which calculates two granulometric curves, one for each mode of the original distribution. So finer and coarse fractions are examined and the respective D50 and D90/D30 are calculated. In this way it's possible to compute the sediment transport respective to each mode of the granulometric curve, and the eroded volume which will modify the grain-size distribution of the breach bed. Until the distribution lasts bimodal, the granulometric curves are again computed, at each time step, using the weights of the remaining sediments; however, when one of the two classes is eroded, in such a way to consider the total granulometric distribution as a unimodal one, the new D50 is calculated and the simulation is resumed with the original program. This last simulation achieved interesting results because it has been possible to simulate the stop of the erosion of the bed of the breach, due to the formation of a unerodible threshold caused by the increase of the relative frequence of the boulders in the breach bed. This phenomenon called armouring (not provided by the original program) makes the new model more realistic and reliable in forecasting the evolution of the breakdown of a landslide dam