Rock Falls and Risk Evaluation

Safety and Security Engineering 675

Rock falls and risk evaluation

M. Papini, L. Longoni & M. Alba Milan Polytechnic, Italy

Abstract

Rock falls are really frequent on all Alpine and Pre-Alpine areas and these landslides have recently been front-page news. Among those, a rock fall involved the surroundings of the tourist resort of Varenna (area of Fiumelatte). The episode resulted in two human victims and seriously damaged some buildings and the railroad Milano-Lecco-Sondrio-Tirano. Because of rock fall diffusion, frequency and randomness, accurate scientific procedures for hazard prediction have recently been developed. Such procedures are focused on the definition of the areas of highest rock-fall hazard, to set up fast safety measures for preventing injuries. Rock-fall hazard evaluation procedures are actually based on mathematical models, which usually try to reproduce the geological phenomena. Such models usually require either the slope topographic characteristics or the rocks geo-mechanical properties. As slight differences in model parameters usually bring notable changes in events prediction, data collection procedure has to be clearly defined. In this paper we will bring focus on the important role that the topographic data collection plays in the accuracy of the previously mentioned models. Thus, we will compare results of risk evaluation procedure carried out using a modern laser scanning technique with results obtained by means of a traditional and widely used cartography technique. Keywords: rock fall, hazard, laser scanning.

1 Introduction

This study is based on observations on a rock fall that involved the Varenna built-up area on November 13th, 2004. The present work is focused on hazard evaluation for similar rock-fall types using data collected in this event. For assessing limitations and advantages of actual prediction methods, different topographic approaches have been considered. Rock-fall hazard papers are based

WIT Transactions on The Built Environment, Vol 82, © 2005 WIT Press www.witpress.com, ISSN 1743-3509 (on-line) 676 Safety and Security Engineering on individuation and delimitation of high-risk areas and phenomena classification is made using the most probable fall-trajectories. Risk magnitude can be estimated using numerical models for rock-fall simulation and the stop zone prediction. Model reliability depends on accuracy of input parameters, especially rocks geo-mechanical properties and slope topographical characteristics. This paper shows the influence of slope topographic measurements and model digital resolution on risk prediction, while geo-mechanical rock properties have been measured on some specimens collected in situ.

2 Rock fall

Rock fall represents one of the highest risks in mountain area. A recent bibliographical study outlines that rock fall is one of the most dangerous phenomenon for people safety. Historical data shows that such events are really common in Lombardy. Such frequentness is mainly due to geomorphologic characteristics, geological local circumstances (essentially lithological- structural), pluviometric phenomenology and climate (characterized by frequent froze-unfroze cycles). According to latest researches the highest risk areas are Sondrio and Lecco province and Garda lake area (INTERREG II C). In Lecco Province rock falls are 28.4% of overall censed landslides while in Sondrio province this percentage rises to 30%. One must notice that risk of infrastructure and built up areas damaging is really high, and local public administrations have to evaluate large area dangerousness. For current purposes, a three dimensional model plays a really important role in risk area forecasting. A rock fall consists in blocks fall due to slope traction or shear joints. Phenomenon can be divided in different steps, that are detachment, free falling, impact, bouncing, and sliding. Number of events that may cause a fall is wide and almost unpredictable. In rock fall study analysis of detachment, free fall and stop is essential. As previously mentioned, detachment is due to a critical combination of intrinsic factors (Clerici 2000).

3 Methods of study

Some authors propose different approaches for rock slopes instability study. This paper will not go into detail on different slope instability prevention methods; however some of the more used analysis techniques will be summarized. Each method makes statements about the rock slope stability by means of some parameters considered more important than others. Due to phenomenon complexity, every model contains some simplifications, although some common characteristics can be found in each model (rock characteristics, joints, rainfalls and other parameters). The INTERREG II C program - Rock slopes instability prevention - indicates two different approaches to the problem: global methods and specific methods. Global Methods are: RES and RHAP, CETE method and MATTEROCK method. Specific Methods are: Ligrim, Probabilistic approach and trajectography methods. This paper will bring focus on the part of the model

WIT Transactions on The Built Environment, Vol 82, © 2005 WIT Press www.witpress.com, ISSN 1743-3509 (on-line) Safety and Security Engineering 677 describing fall trajectories. In this paper Varenna rock fall is used to demonstrate the importance of topography in trajectory lateral dispersion. Different resolution of topographical survey has been used as input for a three dimensional ROTOMAP mathematical model for pointing out an optimal resolution for hazard area definition. Different authors have demonstrated that rock pattern depends on geomorphology and notable errors may occur if the calculus model is not sufficiently accurate. Precise topographical survey usually leads to accurate predictions (Focardi [3]).

4 Survey techniques

Up to now terrain survey has been performed using techniques developed by topography. Latest evolutions, as total stations allow time saving and high accuracy when compared to tachometers. Combining these techniques with photogrammetry permits obtaining accurate results; for example aerial survey can be used to create terrain digital model of wide areas that can be used for slope stability problems. Nevertheless these techniques combination usually requires long lasting post-processing data analysis. Analysis used in this paper are carried out by means of a laser scanner, an experimental technique that allows time saving and high accuracy either in data collecting or processing. Table 1 shows accuracy of different known methods for large scale survey and monitoring (M.Caprioli, G.Strisciuglio, E. Tarantino). As we will later demonstrate, the use of laser scanning can significantly increase the rock trajectory estimation and therefore the hazard area evaluation.

Table 1: Topographic methods.

Method RANGE ACCURACY GPS <20 Km 5-10 mm + 1-2 ppm topographic network <300-1000 m 5-10 mm geometric levelling Variable 0.2-1 mm / Km land photogrammetry <100 m 20 mm da 100 m aerial photogrammetry H<500 m 10 cm land laser scanning Variable 3 mm da 100 m aerial laser scanning variable 10-15 cm

4.1 The laser scanning technique

Laser scanning technique represents an evolution of photogram metric and topographic techniques for digital model acquisition (D.S.M. - Digital Surface Model). Laser scanner does not require any measured surface conditioning if it scatters sufficient energy needed for measuring. The surface measured by a laser scanner is modeled through a discrete group of points. For pointing out whether laser scanning represents the ideal method for rock trajectory prediction or not we will compare data collected using other

WIT Transactions on The Built Environment, Vol 82, © 2005 WIT Press www.witpress.com, ISSN 1743-3509 (on-line) 678 Safety and Security Engineering topographic techniques. In table 2 are listed main advantages and disadvantages of the laser scanning technique.

Table 2: Advantages and disadvantages of the laser scanning technique.

ADVANTAGES DISADVANTAGES High accuracy in the three dimensional Time-expensive post processing image acquisition Geometrical meaningful information Scanner precision indices floor noise in especially when compared to other measurements. Data filtering is often techniques. required for reducing disturbances without loosing terrain details Accurate overlapping between geometrical Laser scanning technology is not mature and data and photographical mapping several aspects have still to be developed. Actually post processing is quite complex because of different tasks are performed by different software and compatibility problems may arise Easy identification and study of objects and Lack of knowledge in post processing areas, either under topological or documentation; usually long working conservative aspects experience is required Three dimensional model can be under Proper laser scanner system choice is not sampled depending on analysis target easy In some particular fields, laser scanner is not the best way for digital surface mapping Instrumentation cost is still really high

Laser scanner can be used in two different configurations: land laser (T.L.S.) and aircraft laser (L.I.D.A.R.). T.L.S is an evolution of the total station and its primary application is the rock slope survey. The instrument is therefore efficient in geometrical digitalization and shows great advantages when compared to topographic and photogrammetric techniques. Compared to topographic techniques it can run automatically and it is therefore more practical, while when compared to photogrammetric techniques it is easier to use.

5 Rock fall in Varenna

This part of the paper models the rock fall that involved the Varenna built up area for evaluating the effectiveness of hazard prediction methods: the results obtained in simulations are compared to real data. As input for the simulation, different survey techniques have been used and their results have been compared for identifying the best survey technique. Area implicated in rock fall was classified as unstable ever since; the territory frequently collapses and slopes above Varenna, Fiumelatte and Pino built up areas are intensely jointed. “At 5 PM, on November 13th, 2004 at about 600m above the sea level from Foppe Mountain, 15000 m3 of rock collapsed towards Fiumelatte, destroying two houses and seriously damaging other five buildings. Furthermore the railway Milano-Lecco-Sondrio-Tirano was interrupted because of rocks on the rails and

WIT Transactions on The Built Environment, Vol 82, © 2005 WIT Press www.witpress.com, ISSN 1743-3509 (on-line) Safety and Security Engineering 679 two electrical power lines have been damaged. The biggest rock (110m3) caused two victims. (Lombardy region report, Zaccaone A. Bonalumi G.).

5.1 Area geological classification

Some authors (Francani V. 1971, Cancelli A. 1987, etc.) have studied this area from the geological point of view. In this paragraph a synthesis of these previous studies are presented. In the studied area sedimentary triassic rocks outcrop: the “Calcare di Perledo-Varenna” and the “Calcare di Esino”. The first consists of black limestone with some flint nodules. These rocks are about 400 m thick on the Grigna area. Particularly in the landslide area thinly bedding black limestone diffusely outcrop. The area is affected by some folds and faults. The Calcare di Esino is etheropic with Calcare di Perledo-Varenna. It consists of grey limestone and dolomites. These sedimentary rocks are affected by close fissuration, tending to divide it into blocks, and by karst phenomena. These rocks form vertical wall where the rock fall has been happened. Some joints and faults subdivide the bedrock in more or less dislocated blocks which are the intrinsic factor of the slope instability. Some debris deposits are localized on the basis of the instability slope They predict that the area is characterized by the old rock falls and by the potential future hazard.

5.2 Rock fall trajectografical study

Knowing the rocks trajectories in Fiumelatte event, we will compare simulations starting from different topographic survey techniques looking for the one that gives the best lateral rock dispersion. Analyses are carried out using software divided into two packages: ISOMAP for 3D Surface Modelling and ROTOMAP for Rockfall Analysis. ROTOMAP uses a statistical approach: it simulates a wide number of physically admissible events and proposes the most probable one; result of the simulation is not a defined point but a “stop probability area”. It is therefore possible to make assessments about risk in the investigated area using statistical evaluation given by ROTOMAP software.

5.2.1 Three dimensional model reconstruction The first fundamental operative part consists in geometrical slope modeling: for this purpose one can use a detailed and up-to-date cartography or a proper topographic survey. In the present paper three different techniques will be analyzed: • CTR techniques; • Aerophotogrammetric techniques; • Laser scanning techniques. Grid is chosen considering the resolution required and time for data input (a grid too thick gives a good detail but requires a lot of time for data input and manipulation). Grid size must be chosen for containing either area that has to be protected or rock fall source areas. Huge grids don’t usually produce benefits and cause a waste of time in data input. When looking for the optimal solution one must consider the number of meshes that ISOMAP can use. Large grid size

WIT Transactions on The Built Environment, Vol 82, © 2005 WIT Press www.witpress.com, ISSN 1743-3509 (on-line) 680 Safety and Security Engineering usually leads to low model resolution while thick grids may create peaks due to measurement noise, and therefore give inadequate results. Model results have been analysed in order to evaluate the sensitivity of rock-fall hazard papers on different topographic survey technique, and not model efficiency. The same geomechanical parameters are used to estimate the comparative analysis of model results. The using software allows two different methods for digital terrain model reconstruction. In this paper limit polynomial surface method was used to perform the simulation. The use of this polynomial algorithm makes it possible to generate points that are external to the area of the sampled locations in such a way as to maintain the trend of the surface, even when this surface cannot be approximated by a simple horizontal plane (ISOMAP & ROTOMAP – Geo&soft Ing. Scioldo). The first stage consists of creating 3D model by ISOMAP. The topographic maps have been generated with different input data points: Using CTR; Using Aerophotogrammetric techniques (fig. 1); Using Laser Scanning techniques (fig. 2). The comparative analysis of the figures 1-2 shows the differences in the accuracy of the topographic models. There are no differences in the 3D Model generating with CTR input data and with Aerophotogrammetric ones.

1522700 1522800 1522900 1523000 1523100 1523200 5094100 5094100

5 0

0 0

0 3 4 6 0 0 0 0 5094000 5094000

0 5093900 5093900

1522700 1522800 1522900 1523000 1523100 1523200

Figure 1: DTM with aerophotogrammetric technique.

1522800 1522900 1523000 1523100 1523200 1523300 5094200 5094200

500 5094100 5094100

0 5094000 5094000

5093900 5 5093900

1522800 1522900 1523000 1523100 1523200 1523300

Figure 2: DTM with laser scanning technique.

Probably it is due to the choice of the grid size that cannot allow assessing the advantages in the use of aerophotogrammetric technique. The topographic map (fig. 2) generated with input data points of laser scanner technique has a lot of differences with the other two plans. The precision in this map will allow better representing the deviation of the falltrack; but this topographic survey technique has a lot of disadvantages especially in the post-processing data. The 4 paragraph also deals with these kinds of problems: a huge amount of data, the efficiency of the instruments, the lack of knowledge in this sort of study etc. Actually post processing is quite complex because of different tasks are performed by different software and compatibility problems may arise. Moreover data filtering is required for reducing floor noise in the measurements and a robust filtering method is necessary for DTM generating from laser scanner data with the presence of vegetation. In spite of filtering the result obtained in the DTM is a good compromised between accuracy and software capability. In fact in the grid parameters of ISOMAP the mesh size should be such as not to exceed 500 grid elements along each axe.

5.2.2 Rock fall analysis The second step consists to analyse rockfall with ROTOMAP module. ROTOMAP is a tree-dimensional model that is used for rockfall analysis. Despite apparently simple, the rockfall problem is quite complex because of the real behaviour of boulders, when rolling, bouncing and sliding, depends on geometric and mechanical details. For geometrical details the previous chapter has already analysed the different technique to improve DTM model. For mechanical details the software required these parameters: normal and tangential energy coefficients of restitution, friction coefficients of the rolling boulders and limit angle. In this study the parameters have been chosen through a back analysis, which is always necessary to calibrate the final model. Despite this back-analysis it’s very important the accuracy of the initial data collection to set up a truly reliable model. In this paper it can be stressed only some information about the parameters: These parameters should be calibrated for the rock boulder size; The limit angle is used to determine when the boulders start to bounce over the slope or return to the rolling conditions. This parameter is the most difficult to evaluate; it depend on rock boulder size and on the micro-topographic conditions (such as slope roughness, vegetation etc.) The kinematics behaviour (travelling modes) depends on friction coefficient of the rolling boulders. In this case the main travelling mode was sliding. After the definition of the geotechnical parameters ROTOMAP can calculated the rockfall paths. The comparative analysis among the three different models (with the three different topographic techniques) shows that the CTR one is the worst one. The simulations based on aerophotogrammetric and laser scanner input data points (fig. 3 – fig. 4) are the best representative for the definition of the falltrack. Here are some differences between the two simulations: ANALYSIS OF ROCKFALL TRAJECTORIES: The rockfall trajectories of the aerophotogrammetric model run on two different channels, while the

rockfall trajectories of the laser scanner model run on one single channel. On November 13th 2004 the rockfall trajectories run on the same channel defined in the simulations calculated through the laser scanner model. The sensitivity of lateral dispersion of the rockfall trajectories is very important in the editing of rockfall hazard paper. ANALYSIS OF THE ROCKFALL RUNOUT ZONES: The rockfall run out zones for the aerophotogrammetric model are all over the Varenna built up area. A comparison of the stopping points during this event with the frequency distribution calculated by the program for the laser scanner model confirms this model like the best one. ANALYSIS OF TRAVELLING MODES: The two models are accurate in the definition of the Kinematical behaviour. The unique travelling mode is the sliding.

Figure 3: Rock fall trajectories of the aerophotogrammetric model.

Figure 4: Rock fall trajectories of the laser scanning model.

In this paper the simulations were made to consider the effects on the numerical results of the different topographic survey technique. The best method

WIT Transactions on The Built Environment, Vol 82, © 2005 WIT Press www.witpress.com, ISSN 1743-3509 (on-line) Safety and Security Engineering 683 to evaluate the rockfall hazard is laser scanner. The figures 5 and 6 show the rockfall run out zones: the first one is for the real event (13 November 2004), the second one is the result of the simulation for the laser scanner model. Different topographic technique could be chosen to perform numerical simulations but only with a detailed description of the micro-topographic conditions it is possible to evaluate the true rockfall hazard.

Figure 5: Run out zones for the real event, Varenna on 13 November 2004.

Figure 6: Run out zones realized with laser scanner data.

6 Conclusions

In this paper we have shown the improvement in risk evaluation using different topographic technique. A comparison of the historical event (on 13 November 2004) with the results of simulations has illustrated the best topographic technique to estimate the run out zones and the trajectories. The performance of the different models has showed a great estimation of trajectories and run out zones only with the use of laser scanner.

The proposed technique has allowed to generate a DTM model with micro- topographic details and also to define the true crown of the landslides. In this event the rock fall sources areas are vertical rock walls without vegetation; these: are the best conditions to apply laser scanner. The use of this kind of investigations allows evaluating the rock fall source areas, the trajectories and the run out zones. Rock fall hazards evaluation procedures have to be based on the research of these three elements. We have analysed the sensitivity of these three elements as a function of DTM resolution. Both lateral dispersion of trajectories and kinematics variables of simulations strongly depend on the quality of elevation data available for the generation of the topographic model. Anyway, it can be stressed that the values usually assumed for the coefficients of restitution and the friction coefficient contain implicit information about local micro-topographic conditions (slope roughness, vegetation, etc.) and/or block geometry. As a consequence, different values could be chosen to perform numerical simulations and to calibrate models when topographic descriptions of different detail and scale are available (Crosta and Agliardi [3]). Although apparently these coefficients can improve the performance of the simulation it would be even better to use a topographical survey specifically made for the purpose of the study. The more detailed maps generating with laser scanning technique show this is the better criteria to assess rock fall hazard and the associated risk. Despite the problems in post processing data of laser scanning, the resolution of DTM used in this paper is a good compromised between accuracy and computing time.

References

[1] Buretta, P. and Santo, A., Morphostructural evolution and related kinematics of rockfalls in Campania: a case study. Engineering Geology 36, 197-210. [2] Baillifard, F., Jaboyedoff M. and Sartori M., Rockfall hazard mapping along a mountainous road in Switzerland using GIS-based parameter rating approach. Natural Hazards and Earth Sciences (2003) 3: 431-438. [3] Crosta G.B. and Agliardi F., Parametric evaluation of 3D dispersion of rockfall trajectories. Natural Hazards and Earth System Sciences (2004) 4: 583-598. [4] Focardi P., Considerazioni cinematiche sul percorso di massi provenienti da frane di crollo. Geologia Tecnica (1982) 4: 13-23. [5] Guzzetti F., Carrara A., Cardinali, M., and Reichenbach P., Landslide hazard evaluation: a review of current techniques and their application in a multi- scale study, Central Italy, Geomorphology, 31, 181-216, 1999. [6] Guzzetti, F. and Crosta, G., Programma Stone: Un programma per la simulazione tridimensionale delle cadute massi, in: Programme Interreg IIc – “Falaises”, Prévention des mouvements de versants et des instabilités de falaises, 70-79, 2001. [7] Hoek E., and Bray J., Rock slope engineering, revised third edition, London, 358, 1981.

[8] Jaboyedoff M., Baillifard F., Philippossian F. and Rouller J-D., Assessing fracture occurrence using “weighted fracturing density”: a step towards estimating rock instability hazard. Natural Hazards and Earth System Sciences (2004) 4: 83-93. [9] Montgomery D. R., Dietrich W. E., and Sullivan K., The role of GIS in watershed analysis, in: Landform monitoring, modelling and analysis, edited by Lane, S.N., Richards, K. S., and Chandler, J.H., J.Wiley and Sons, New- York, 241-262, 1998. [10] Paar G., Nauschnegg B., and Ullrich A., Laser scanner monitoring – technical concepts, possibilities and limits. Natural Hazards Workshop, 5-7 June 2000, Austria.