Field and monitoring data of debris-flow events in the Swiss Alps

Transcription

1 Field and monitoring data of debris-flow events in the Swiss Alps M. Hürlimann, D. Rickenmann and C. Graf 161 Abstract: Debris flow is a common process in the Swiss Alps and in other mountainous parts of the world. The understanding of debris-flow behaviour is essential to assess the hazards they present. An important approach towards improving the knowledge of debris-flow processes is the gathering of real-time data by debris-flow observation stations. Observation stations were established in three Swiss debris flow prone watersheds and equipped with video cameras, ultrasonic devices, a radar device, geophones, and rain gauges. In 2000, four significant debris flows were observed. The data provided useful information on the mechanics of debris flows and on the efficiency of the measuring devices. The observed debris flows are characterized by volumes between and m 3, front velocities ranging from 2 to 5 m/s, and peak discharges between 20 and 125 m 3 /s. The analysis of the monitoring data revealed that ultrasonic and radar devices are very helpful tools, whereas the quality of the geophone signal strongly depends on the substrate on which the instrument is installed (i.e., bedrock versus unconsolidated material). Video images are useful to verify the data obtained by the other devices. A dynamic analysis of one debris flow was carried out and the simulated results are in fair agreement with the observed data. Key words: debris flow, Swiss Alps, monitoring, dynamic analysis. Résumé : Les coulées de débris sont un processus commun dans les Alpes suisses et dans les autres régions montagneuses du monde. La compréhension du comportement des coulées de débris est essentielle pour évaluer les risques qu ils présentent. Une approche importante pour améliorer la connaissance des processus des coulées de débris est l accumulation de données en temps réel par des stations d observation des coulées de débris. Des stations d observation ont été établies dans trois bassins versants suisses sujets à des coulées de débris et ont été équipées avec des vidéos caméras, des dispositifs ultrasoniques, un appareil radar, des géophones et des jauges de pluie. Les données ont fourni des informations utiles sur le mécanisme des coulées de débris et sur l efficacité des appareils de mesure. Les coulées de débris observées sont caractérisées par des volumes entre et m 3, des vélocités de la face frontale de 2à5m/setdesdébits de pic entre 20 et 125 m 3 /s. L analyse des données de mesures révèlent que les dispositifs ultrasoniques et les appareils radars sont des outils très utiles, alors que la qualité des signaux des géophones dépendent fortement du substrat sur lequel l instrument est installé (i.e., lit rocheux vs matériau non consolidé). Les images vidéo sont utiles pour vérifier les données obtenues par les autres mécanismes. Une analyse dynamique d une coulée de débris a été réalisée et les résultats simulés sont en concordance raisonnable avec les données observées. Mots clés : coulée de débris, Alpes suisses, instrumentation, analyse dynamique. [Traduit par la Rédaction] Hürlimann et al. 175 Introduction Debris flows are rapid, gravity-induced mass movements consisting of a body of granular solids, water, and air (Varnes 1978). They form a severe natural hazard in mountainous regions due to their high velocity, large volumes, and frequent recurrence in a wide spectrum of morphologic settings. The causes of debris flows can be different but a frequently postulated triggering mechanism is the influence of water and the resulting increase of pore-water pressure. A debris-flow event often includes a series of surges characterized by maximum flow depth and peak discharge. Limited data are available on the real-time monitoring of debris flows. Many detailed field observations of debris flows have been made in Japan (Suwa 1989) and in China (Zhang 1993). More recently, automatic observations on debris flows have also been gathered in Italy (Arattano et al. 1997; Berti et al. 1999; Arattano and Marchi 2000; Berti et Received 3 January Accepted 22 August Published on the NRC Research Press Web site at on 27 January M. Hürlimann, 1,2 D. Rickenmann, 3 and C. Graf. WSL, Swiss Federal Research Institute, Zürcherstrasse 111, 8903 Birmensdorf, Switzerland. 1 Present address: Department of Geotechnical Engineering and Geosciences, Technical University of Catalonia (UPC), Jordi Girona 1 3 (D2), Barcelona, Spain. 2 Corresponding author ( marcel.hurlimann@upc.es). 3 Present address: Department of Forest and Mountain Risk Engineering, University of Agricultural Science, Peter Jordan-Straβe 82, 1190 Wien, Austria. Can. Geotech. J. 40: (2003) doi: /T02-087

2 162 Can. Geotech. J. Vol. 40, 2003 al. 2000). There are still uncertainties in scaling findings from laboratory studies to prototype conditions. Therefore, direct observations of debris-flow parameters in the field are of great importance. Moreover, accurate field data are essential for the back calculation of debris flows by numerical simulation models. Debris flows are common processes in the Swiss Alps and several events have produced significant damage and fatalities during recent decades (e.g., Rickenmann and Zimmermann 1993; Zimmermann et al. 1997a). After the catastrophic debris flows of 1987, a comprehensive study was carried out on debris-flow type, source area geology, debris-flow volume, and rainfall thresholds for debris-flow initiation (VAW 1992). In recent years, research has focused on laboratory experiments and numerical simulations (Rickenmann and Koch 1997: Rickenmann and Weber 2000; Tognacca 2000). Furthermore, three debris-flow observation stations were installed in Switzerland. Two of them (Dorfbach and Schipfenbach) were completed in the summer of 1997, whereas the third observation station (Illbach) was installed in 2000, see Fig. 1. In this paper, we present field and monitoring data of four debris flows that occurred during 2000 at the observation stations. The events took place in the Schipfenbach and Illbach catchments. No debris flow occurred in the Dorfbach. Initially, we discuss the debris-flow event at the Schipfenbach and analyze the data. Dynamic analysis of this debris flow was carried out, and simulated results were compared with the field and monitoring data. Then, we present data on the three debris-flow events that occurred at the Illbach. The objective of this paper is to analyze the characteristics of the debris flows that occurred at Schipfenbach and Illbach in The rainfall data provide useful information on the initiation of the debris flow, whereas field and monitoring data improve the understanding of the flow behaviour. This work will be helpful in calibrating numerical models against field and monitoring data. General setting The locations of the three debris-flow observation stations are illustrated in Fig. 1. The observation station at Schipfenbach near Silenen is situated in the central part of Switzerland in the northern Alps. The other two sites are installed at the Dorfbach near Randa and at the Illbach near Susten. Both are located in the southern part of Switzerland where the large alpine valley of the Rhone River is the main morphologic feature. The three catchments were selected out of many possible sites in the Swiss Alps because of persistent debris-flow activity and good accessibility. The Illbach, for example, is known as the most active debris-flow system in the Swiss Alps. The Schipfenbach catchment The topographic features of the Schipfenbach catchment are illustrated in Fig. 2a and some general data on the drainage basin are listed in Table 1. The highest point of the drainage basin is the Läged Windgällen peak, while the torrent flows into the Reuss River at the lowest point of the debris fan. A debris retention basin was built between 1986 Fig. 1. Location of the three debris-flow observation stations in Switzerland. Table 1. Morphology of the Schipfenbach and the Illbach drainage basins. Schipfenbach Illbach Area (km 2 ) Rocky area (%) Forest (%) Grassland (%) Lake (%) 2 Maximum elevation (m a.s.l.) Minimum elevation (m a.s.l.) Exposure W N Mean slope angle of channel (%) Mean slope angle of fan (%) and 1988 in the upper part of the debris fan at ~650 m a.s.l. The area of the watershed upstream of the retention basin is 1.8 km 2. The main geologic unit of the catchment is limestone with less abundant granitic rocks and gneisses. In 1936, a series of rockfalls occurred at the southwestern flank of the Läged Windgällen and covered the upper parts of the drainage basin with a total volume of about m 3. The Schipfenbach catchment can be divided into the following four morphologic zones. (1) The highest parts are characterized by rocky sequences with some high, sub-vertical walls forming the southwestern flank of the Läged Windgällen. (2) Below this zone, unconsolidated debris material, sand to boulder sized, has accumulated as large talus slopes down to ~1100 m a.s.l. In this zone, the torrent is deeply incised into these deposits and several gullies have formed. (3) The zone between 1100 and 780 m a.s.l. is characterized by steep slopes and the torrent bed is underlain by bedrock. (4) A debris fan has formed at the mouth of the drainage basin between 490 and 680 m a.s.l. In the upper part of the debris fan upstream of the retention basin, the torrent is deeply incised into the debris material, whereas an artificial

3 Hürlimann et al. 163 Fig. 2. Topographic map of the (a) Schipfenbach and (b) Illbach catchments indicating the measuring devices. The contour interval is 100 m. channel was built downstream of the basin draining into the Reuss river. Characteristics of the particle size of the debris-flow material in the Schipfenbach can be illustrated by grain size distribution curves. Four samples were analyzed using dry sieving for diameters larger than 4 mm, wet sieving for diameters between 1 and 4 mm, and by sedimentation for diameters less than 1 mm. The material was collected in the debris retention basin on four different dates between 1996 and The results (Fig. 3) show that the gravel fraction of the Schipfenbach material is rather high (45 70%), whereas the clay fraction is very low (<5%). The uniformity coefficient, as large as 825, is characteristic of a very wellgraded material. Climate conditions of the Schipfenbach catchment are influenced by its location at the northern flank of the Central Alps. Winter precipitation is mainly snowfall and heavy rainstorms are common in summer. The mean annual precipitation at the meteorological station of Altdorf (located about 10 km north of the study area) is about 1310 mm, whereas the mean annual precipitation of the Schipfenbach catchment is approximately 2300 mm due to the orographic effect (Swiss Federal Office for Water and Geology 1999). Historical records on debris flows at Schipfenbach exist back to the 19th century (Fig. 4). Generally, small-scale debris flows (volume less than m 3 ) occur about every 2 years on average, while large events with volumes of about m 3 or more occurred only three times in the 20th century (1929, 1977, 1987). Debris-flow activity increased after the rockfalls in 1936, but large events did not occur. The Illbach catchment The topographic features of the Illbach catchment are illustrated in Fig. 2b and some general data on the drainage basin are listed in Table 1. The watershed has an area of 10.5 km 2 and an exposition to the north. The highest point of the basin is the Illhorn and the Illbach enters into the Rhone River at 610 m a.s.l. Several types of countermeasures were

4 164 Can. Geotech. J. Vol. 40, 2003 Fig. 3. Grain size distributions of debris-flow material collected at the Schipfenbach and the Illbach. Fig. 4. Historical records of debris flows during the 20th century at Schipfenbach and Illbach. The debris flows that occurred during the night of October 2000 at Illbach are illustrated as one event. constructed along the torrent. A large debris retention dam with a height of almost 50 m was built in the southern channel at about 1000 m a.s.l. between 1967 and Downstream of this large dam the torrent is protected by a series of smaller check dams down to the Rhone River. The bedrock in the drainage basin is mainly composed of quartzites, calcareous deposits, and dolomites. Erosion strongly affects the southwestern part where calcareous deposits and dolomites crop out in the steep side walls. The dolomite is unusually susceptible to weathering and provides a large amount of silty material. The calcareous deposits and dolomites are strongly jointed and repeatedly cause landslides. In 1961, a large rockfall occurred with a total volume of about to m 3 in this area of the drainage basin. The loose deposits of the 1961 rockfall and the continuous slope movements are an abundant supply source for the debris flows, which generally consist of a muddy slurry (dolomites) and boulders of quartzite or calcite. The morphology of the Illbach catchment is characterized by two different zones. The northern zone shows no debrisflow activity, the southern zone is a large headwater catchment with intensive erosion, which is the initiation zone for the debris flows. Erosion has resulted in steep, unstable rockfaces that supply the channel with debris. Another characteristic of the Illbach catchment is that the debris fan is unusually large in comparison with similar sized catchments in the region. The fan has a radius of about 2 km and a volume of about m 3 (GEO7 2001). Seven grain size samples from the Illbach were analyzed; four were collected in the creek channel and three were extracted from debris-flow deposits. The results (Fig. 3) show that the Illbach samples contain a larger proportion of fine particles, particularly sand (35 40%), than the Schipfenbach samples (20 25%). The clay fraction is similar to the Schipfenbach s (<5%), and the uniformity coefficient is large with values of up to 225. The climate of the Illbach catchment is strongly influenced by its location in an interalpine valley, which is manifested by a mild climate and a low annual precipitation. The mean annual precipitation ranges from 700 mm in the lower part of the drainage basin to 1700 mm at the summit region (Swiss Federal Office for Water and Geology 1999). Intense rainstorms occur mainly in summer, and estimated rainfall intensities are between 35 and 57 mm/h for a 0.5 and 1 h rainfall duration, respectively, corresponding to a return period of 100 years (Swiss Federal Office for Water and Geology 1999). Historical data on the debris-flow activity in the Illbach exist from the beginning of the 20th century (Fig. 4). The debris-flow chronology shows that debris flows have oc-

5 Hürlimann et al. 165 Fig. 5. Video frames of the Schipfenbach debris flow and the three events recorded at Illbach. For location of video cameras (VC) see Fig. 2. (a) First surge of the Schipfenbach event recorded by VC1. (b) First surge of the Schipfenbach event entering the retention basin, recorded by VC2. (c) Second surge of the Schipfenbach event entering the retention basin, recorded by VC2. (d) Illbach debris flows on (e) 3 June, (f) 28 June, and (g) 24 July. curred regularly during the last 100 years, including many events with volumes less than m 3, five events with volumes ranging from to m 3 and one event in 1961 with a total volume of about m 3 (GEO7 2001). As in the Schipfenbach catchment, a large rockfall event (1961) significantly increased the debris-flow activity in the following years. In the 1970s, the debris-flow frequency decreased on the fan due to the construction of a solid gravity dam in the middle reach of the southern torrent. By the early 1980s this ~50 m high sediment retention dam had filled and is now unable to contain further debris. Consequently, debris-flow activity has increased during the last 20 years. The monitoring system The monitoring systems were installed in early summer 1997 (Schipfenbach) and in 2000 (Illbach) and include video cameras, ultrasonic devices, a radar device, geophones, and rain gauges. The ultrasonic devices are of the type VEGASON Series 70 (VEGA 2001), the radar device is a VEGAPLUS Series 50 (VEGA 1999), the geophones are GS 20 DX (GEO SPACE 1997), and the rain gauges are of the type Lamprecht (LAMBRECHT 2000). The locations of the measuring devices in the Schipfenbach catchment are illustrated in Fig. 2a. Two video cameras (VC1 and VC2) and spotlights were installed near the sediment retention basin. One camera (VC1) provides motion pictures of debris flows in the reach just upstream of the retention basin allowing for analysis of the flow behaviour in the torrent channel (Fig. 5a). The other camera (VC2) supplies pictures of debris flows entering the retention basin and provides information on deposition processes of debris flows and the effect of the retention basin (Figs. 5b and 5c). At 140 and 210 m upstream of the retention basin, two ultrasonic devices (USD1 and USD2) were set up at two cross sections to measure the flow depth and provide debris flow hydrographs. Four geophones (Geo1 Geo4) were installed in the higher part of the drainage basin (Fig. 2a). The geophones are located at elevations between ~1060 and ~1180 m along a total distance of 270 m. Geo3 and Geo4 were installed on bedrock adjacent to the channel, whereas Geo1 and Geo2 were mounted on large blocks situated on the debris-flow levees near the channel. The geophones measure the ground vibration produced by a passing debris flow and log the data as impulses if the amplitude of the vibration, transmitted as voltage, exceeds a threshold magnitude of 200 mv. Thus, the geophone data represent the vibration intensity as integrated information since the sensors record the number of exceedings during 1 s. Moreover, the geophones trigger the measuring devices further downstream during a debris-flow event. The two ultrasonic devices and the four geophones are synchronized allowing calculation of average debris-flow velocity along the channel reach. A rain gauge was installed in the drainage basin at an elevation of 1070 m and records rainfall every 10 min. The measuring devices in the Illbach were installed in 2000 before the onset of the debris-flow season. Three rain

6 166 Can. Geotech. J. Vol. 40, 2003 Fig. 6. Photograph of the material deposited in the retention basin due to the Schipfenbach debris flow. Table 2. Magnitude, flow depth, and velocity of debris flows at Schipfenbach and Illbach in the year Date Volume (m 3 ) Maximum discharge (m 3 /s) Maximum flow depth (m) Schipfenbach 1st surge 6 Aug * 1.2* 3.4* 2nd surge 1.6* 4.6* Illbach 3 June Illbach 28 June Illbach 1st surge 24 July nd surge *Calculated from hydrographs measured by ultrasonic devices (location USD in Fig. 2a). Calculated from hydrographs measured by radar device (location RD in Fig. 2b). Calculated from video camera data (location VC in Fig. 2b). Velocity (m/s) gauges were set up in spring Therefore, the analysis of the debris-flow events in 2000 does not include precipitation data. Three geophones, one radar device, one video camera, and three rain gauges form the monitoring network in the Illbach catchment (Fig. 2b). Except for the rain gauges, all of the devices are situated along the channel on the debris fan. The video camera is located in the lower part of the fan and provides motion pictures of the debris flows along a channel reach of about 100 m length as well as the drop over a concrete check dam (Figs. 5d, 5e, and 5f). The radar device was mounted at the road bridge at 620 m a.s.l., and measures the flow depth of the debris flows before they enter the Rhone River. The radar device provides hydrographs of the debris flows at this cross section in the same way as the ultrasonic devices in the Schipfenbach. The three geophones (Geo1 to Geo3) were installed in the middle and lower part of the fan within about 500 m. The geophones trigger the measuring devices and velocity can be calculated from the time difference between the geophones and the radar device. The three rain gauges are located in the southern part of the drainage basin, which is the primary debris-flows initiation zone. The Schipfenbach event General observation The first monitored debris-flow event occurred in the morning of 6 August 2000 after intense precipitation during the previous 24 h. The geophones triggered the measuring devices at 6:43 am and the beam lights turned on for improved video images. Analysis of the video tapes showed that the debris-flow descended in two main surges and several secondary surges. Subsequent intense sediment transport (hyperconcentrated flow) continued for several hours until the afternoon. The debris flow was stopped in the sediment retention basin and did not cause any damage (Fig. 6). Field data Several field surveys were carried out to determine the most important characteristics of the debris flow. Information on the starting zone, erosion, and deposition along the flow path was of major interest to investigate triggering mechanisms and to estimate the total volume of the event. Data on several cross-sections were collected to analyze the flow behaviour and to determine the evolution of the sediment volume along the debris flow path. The first survey was carried out two days after the debris flow event, two additional field surveys took place about one week later. Magnitude, flow depth, and flow velocity data of the debris-flow event are listed in Table 2. A longitudinal profile of the debris flow path is provided in Fig. 7 including the slope angle and some information on the debris flow. The total length of the debris flow path is approximately 1800 m. The debris flow path can be separated into four different reaches: initiation, erosion, transit, and deposition. (1) The starting zone of the debris flow is located between 1645 and 1660 m a.s.l. in the channel immediately below a rock face in the highest part of the drainage basin (Fig. 7). Significant vertical erosion in this channel reach suggests that the debris flow may have been initiated by bed fluidization or by intense bedload transport transforming rapidly into a debris flow due to high solid concentration at the steep channel gradient. Thus, the initiation mechanism can be defined as in-channel mobilization. Over a 400 m long reach downstream of the initiation zone, the channel

7 Hürlimann et al. 167 Fig. 7. Longitudinal profile and slope angle of the Schipfenbach. traverses scree, and several secondary starting and deposition zones were found within this reach. (2) The 600 m long reach between 1450 and 1100 m a.s.l. is characterized by significant erosion. The debris-flow volume increased considerably due to basal and lateral erosion. Comparisons between pre- and post-event photographs indicated that basal erosion especially increased the debris flow volume. (3) The section below 1100 m a.s.l. can be defined as the transit zone. The debris flow passed this section with little mass change. In the steep reach down to ~780 m a.s.l., the debris flow incorporated only small amounts of loose material due to the rocky characteristics of the channel bed and thus limited amounts of channel fill. However, some major erosion occurred on the highest part of the debris fan ( m a.s.l.). (4) The debris flow finally stopped inside the debris-flow retention basin at an elevation of 650 m, and only a very small amount of fine material passed the basin to enter Reuss River. During the field surveys, flow depth, cross-section area, and the slope angle were measured at 17 different cross sections between 660 and 1605 m a.s.l. The flow depth ranged from 1.3 to 2.8 m depending on the channel width and the flow section areas varied between 4.5 and 23 m 2 depending on the location of the cross section along the debris flow path. In addition, debris-flow velocity was estimated using the superelevation approach (e.g., Johnson and Rodine 1984). This approach was applied in many debris-flow studies and provides the velocity using the following expression [1] v =(gr c cos δtan β) 0.5 where g is the gravitational acceleration, r c is the radius of curvature of the centre-line of the first bend in the channel, δ is the slope angle of the channel bed, and β is the superelevation angle. Recent studies, however, have shown, that these velocity estimates are usually too high (e.g., Jakob et al. 1997). In the Schipfenbach, the lack of clear lateral deposits on both sides of the curves only enabled a confident estimate at two different locations (970 and 1295 m a.s.l.). Assuming an error of about 30% in the estimate of r c and of about 20% for tanβ, the total error of the velocity calculated by eq. [1] is about 25%. The morphology of the cross section at an elevation of ~1200 m was compared with photos and rough measurements available for the years 1971 and 1987 and shows a significant erosion of the torrent channel into the unconsolidated scree due to fluvial processes and debris flows. A maximum erosion depth of about 4 m can be observed at this location. Moreover, erosion rates of 2 3 m were estimated in the same part of the channel comparing photographs of the 1970s with recent ones. The total volume of the debris flow event was estimated using three different methods: (i) topographic survey of the accumulated material in the sediment retention basin (~6000 m 3 ); (ii) calculation of the erosion and deposition along the debris flow path during the field surveys (~4900 m 3 ); and (iii) hydraulic analysis of the ultrasonic device data ( m 3 ). The 6000 m 3 obtained by the topographic survey corresponds to the total volume accumulated in the retention basin and is a maximum value since it includes not only the main surges but also the material deposited by the hyperconcentrated flow following these surges. We suggest that the total volume of the debris flow can be assumed to be m 3. During the field survey carried out 2 days after the event, three samples of the debris flow deposit were collected in the sediment retention basin. The samples had volumes between 1000 and 2500 cm 3 and were extracted form the material of the main debris-flow surges. In the laboratory, selected physical properties of the samples were determined. The natural density of the samples ranged from 2200 to 2400 kg/m 3, the dry density was between 2000 and 2240 kg/m 3, and the water content ranged between 8 and 10% by weight. Monitoring data The rain gauge, the geophones, and the ultrasonic devices provided important data to enable a detailed interpretation of the debris-flow event. In this section, the most important results obtained from the analysis of the monitoring data will be summarized. The precipitation data were obtained from the rain gauge situated in the drainage basin of the Schipfenbach catchment at an elevation of about 1070 m. Figure 8a shows the daily precipitation during 1 month prior to the debris-flow event. After a rather dry June, July 2000 was characterized by a high rainfall amount of about 189 mm. However, the

8 168 Can. Geotech. J. Vol. 40, 2003 Fig. 8. Precipitation analysis of the Schipfenbach debris flow. (a) Cumulative rainfall during the 24 h prior to the debris-flow event (arrow indicates the time of initiation). Inset shows the daily precipitation during the month prior to the debris flow. (b) Comparison between the climatic threshold for debris-flows initiation in the outer parts of the Swiss Alps and the data for the Schipfenbach event. precipitation amount was not exceptional until the day of the debris flow (6 August 2000). The 24 h antecedent rainfall before the debris flow was 56 mm (Fig. 8a) and until 3:00 am of 6 August, the antecedent rainfall was about 20 mm. From 3:00 am to the initiation of the debris flow at 6:43 am, intense rainstorms occurred in the area of the Schipfenbach catchment. Maximum rainfall intensities reached 11 mm/h during the 3 h period prior to the debris flow. The rainstorms lasted until the afternoon and a total rainfall amount of ~106 mm was measured at the rain gauge for 6 August. These rainfall data can be incorporated into various empirical approaches referring to threshold levels of debrisflow triggering (e.g., Caine 1980; Wilson and Wieczorek 1995). Zimmermann et al. (1997b) determined a relationship between rainfall intensity and duration for debris-flow initiations in the subalpine zones of Switzerland. They proposed the following equation to define a lower limit of debris flow initiation: [2] I =32D 0.70 where I is the average rainfall intensity in mm/h and D is the rainfall duration (in h). A comparison of this threshold with the rainfall conditions of the debris-flow event in the Schipfenbach catchment is shown in Fig. 8b. Rainfall conditions in the Schipfenbach were consistently below the threshold defined by eq. [2]. For this particular event, a period of about 4 24 h prior to the event appears to have been most critical, i.e., the conditions were closest to the threshold defined by eq. [2]. However, there is some uncertainty concerning eq. [2] because the rainfall data used for its development do not necessarily represent the exact locations of debris-flow occurrence. Furthermore, the exact time of occurrence is often unknown. The geophones began recording during the passage of the first main surge at 6:43 am. Figure 9 illustrates the data of the second main surge that were measured by the four geophones (Fig. 2a). The time of the assumed passage of the debris-flow front is indicated for each geophone in the graph. These times were used for the velocity estimates. The two geophones mounted on boulders in the unconsolidated levee deposits (Geo1 and Geo2) recorded only rather weak data during the passage of the debris flow (few impulses in Fig. 9). In contrast, the two geophones installed on bedrock (Geo 3 and Geo4) collected good data with strong peaks of the debris-flow front (many impulses in Fig. 9). The results

9 Hürlimann et al. 169 Fig. 9. Geophone data of the Schipfenbach debris flow. The arrows indicate the front of the second debris-flow surge as it passed each of the four geophone locations. Fig. 10. Hydrographs of the Schipfenbach debris-flow event obtained from the two ultrasonic devices. The fronts of each debris-flow surge are clearly visible. illustrate that the accuracy of the geophone data strongly depends on the subsurface (bedrock versus unconsolidated material) at the sensor location. The data of the two ultrasonic devices USD1 and USD2 were used to plot hydrographs of the debris-flow event (Fig. 10). The debris-flow event included two main surges separated by about 2 min, and several secondary surges. The first main surge seems to have been of lower magnitude than the second one. The mean flow depth is 1.2 m for the first surge and 1.6 m for the second surge. However, significant channel bed erosion occurred during the debris flow and true flow depths can only be assumed. After the debris flow event, the channel bed erosion continued and a final erosion depth of about 50 cm was measured at the cross sections of both ultrasonic devices. Thus, maximum flow depth may have been higher than given above due to the channel bed erosion down to an unknown depth during the debris flow. The most important results obtained from the field and monitoring data are summarized in Table 3. Velocity estimates range from 3.2 to 9.3 m/s, cross sectional areas from 6.75 to 20 m 2, and flow depths from 1.4 to 2.5 m. The peak discharges calculated from velocity and cross sectional areas may have exceeded 80 m 3 /s. From field observations for the data the Chezy coefficients, C, were estimated assuming quasi-steady and quasi-uniform flow conditions [3] C ν ( h sin δ) / 12 where v is the velocity, h is the flow depth, and δ is the slope angle of the channel bed. The values estimated by eq. [3] for the Schipfenbach event are in the range of C = 4 9 m 0.5 /s. The data in Table 3 were compared with the flow behaviour simulated by a numerical model. Dynamic analysis of flow behaviour To model the debris flows hydrographs, the onedimensional case and the approximation of a homogeneous

10 170 Can. Geotech. J. Vol. 40, 2003 fluid was considered (e.g., MacArthur and Schamber 1986; Takahashi 1991; Choi and Garcia 1993; Laigle and Coussot 1993; Hungr 1995; Costa 1997; Imran et al. 2001). Several simulation models were formulated, including a Newtonian turbulent flow resistance term (similar to that used in hydraulic simulations of shallow water flows), and these models were applied with some success to describe the movement of debris flows or lahars (Weir 1982; Arattano and Savage 1994; Hunt 1994; Shieh et al. 1996; Costa 1997). More sophisticated numerical simulation models consider the two phases (water and solids) separately (e.g., Takahashi et al. 1992) and also distinguish between different flow regimes (Nakagawa et al. 2000; Nakagawa and Takahashi 1997). Other model formulations also consider pressure variations within the mixture (Iverson et al. 2000) and account for solid and fluid stresses separately (Iverson and Denlinger 2001). An important advantage of these models is that they allow for variations of the mixture properties along the surge of a debris flow, e.g., the decrease in solids concentration and in the proportion of coarse particles from front to tail. At present, sophisticated debris-flow simulation models assume values for many of the model parameters. In this study a relatively simple approach was used to replicate the observed flow behaviour in Schipfenbach. In an earlier study, we compared several flow resistance approaches implemented in a numerical simulation model and applied them to debris flows at the Kamikamihori field site in Japan (Rickenmann and Koch 1997). We tested different simple rheological models, such as the Bingham fluid, a Newtonian laminar fluid, a dilatant grain shearing model, a Newtonian turbulent fluid, and a Voellmy fluid. We found best agreement of the observed global flow behaviour using the latter two approaches. A similar conclusion concerning the estimation of debris-flow velocity is obtained from the analysis of a large number of field data (Rickenmann 1999). Recently, the Voellmy fluid option of Hungr s (1995) model satisfactorily replicated the runout distance and peak discharge of a large debris flow at Hummingbird Creek in southern British Columbia (Jakob et al. 2000). To simulate the debris flow in Schipfenbach, we used the code AVAL-1D, a one-dimensional model originally developed for snow avalanche analysis. Details on the mathematical formulation and the numerical solution are given in Sartoris and Bartelt (2000), and applications of the model to snow avalanches are described in Bartelt et al. (1999). The model considers the motion of a homogeneous fluid. In this Voellmy two-parameter rheological model the basal shear stress (mainly controlling the flow resistance and the depositional behaviour) consists of a turbulent Chezy-like friction term and a dry Coulomb-like friction, and the total friction slope S f is given as ν [4] Sf = µ cos δ + 2 C R 2 where µ is the dry friction coefficient, and R is the hydraulic radius. Equation [4] is part of the shallow water equations, which describe the unsteady motion of the flowing mixture. Apart from the total mixture volume, the two parameters µ and C mainly control flow depth and velocity. The more fluid-like flow behaviour (at higher velocities) is represented by the Chezy coefficient C, and the more solid-like behaviour (at lower velocities) by the friction coefficient µ. Anadditional component in the model allows active and passive earth-pressure effects to be accounted for (e.g., Hungr 1995). For a number of snow avalanche simulations, inclusion of these effects appears to have a limited influence on the front velocity and the maximum flow depth (Bartelt et al. 1999). Hungr (1995) showed that earth-pressure effects can be important in the case of an earth-flow running up on a counterslope. In the Schipfenbach simulation, we did not include any earth-pressure effects (the ratio of the active and passive coefficients was set to 1). To simulate the debris flow in the Schipfenbach of 6 August 2000 a total volume of 4900 m 3 was released from rest within a reach of about 700 m length, at an elevation between 1700 and 1400 m. In the model, a rectangular channel with a constant width of 7mwasused. This value corresponds to the average top width of the observed flow crosssection areas, which have a trapezoidal shape. The lower flow depths in the computational rectangular sections than in the natural flow cross-sections are roughly equivalent to the hydraulic radius and thus account for sidewall friction. The sediment retention basin is represented in the model with a bedslope of 2.9% and by an increase in width from 7 to 30 m over a length of 90 m, a constant width of 30 m over the next 30 m length, followed by a decrease to 5 m width within the last 20 m of the basin. The concrete outlet structure has a slit opening of 5 m. Downstream, an artificial channel also has a5mwidth. For the numerical simulations, the length of each computational segment was 50 m, except for the sediment retention basin, which had a segment length of 10 m. Simulations were carried out to back calculate the optimal parameter combinations of C and µ. The following criteria were used to evaluate the model results: representation of front velocities and flow cross-sections, as well as runout distance, i.e., deposition of the bulk part of the flow in the sediment retention basin. Figure 11 shows the results of a simulation that produced a reasonable agreement between observations and calculations. Here, the following parameters were used: C =11m 0.5 /s and µ = Upstream of the sediment retention basin (which is located at a distance of 2000 m from the origin) the variations in calculated front velocities represent the variations in bedslope (since the channel width is constant). The simulated velocities agree reasonably well with the observed values. Those marked A, B, E, F, G in Fig. 11 and Table 3 have a higher uncertainty than the other velocity estimates, which were obtained by ultrasonic devices and video frames. The data of A and B were estimated with the superelevation approach explained in eq. [1]. The data of E and F (Fig. 11 and Table 3) refer to short reaches where uncertainties in the interpretation of geophone signals are of importance. A rather slow velocity of 3.2 m/s was calculated for the reach between the lowest geophone and the higher ultrasonic device (Geo4 to USD1, label G in Fig. 11). This reach is characterized by a high mean bedslope of about 60% and several vertical walls with heights of more than 10 m. A possible explanation of the slow velocity in this reach is that the debris flow may have been slowed at the foot of the cascades. Such effects cannot

11 Hürlimann et al. 171 Fig. 11. Results of the dynamic analysis. (a) Longitudinal profile and channel width used in the simulation. (b) Simulation results for C =11m 0.5 /s and µ = 0.175, showing the variation of the front velocity (solid line) and the maximum flow area (dashed line) along the debris-flow path. Labelled thick lines and squares represent velocity values obtained from field and monitoring data (as listed in Table 3). Dots indicate the estimates of cross-sectional areas observed in the field. Table 3. Front velocities and cross-sectional areas of the main surge of the Schipfenbach event achieved by field and monitoring data. Label* Method Length of reach (m) Mean bedslope (%) Estimated velocity (m/s) Cross-sectional area (m 2 ) Maximum flow depth (m) A Superelevation B Superelevation C Geo1 Geo D Geo1 Geo E Geo2 Geo F Geo3 Geo G Geo4 USD H USD1 USD I Video analysis *Labels as indicated in Fig. 11. Field data. Calculated by eq. [3]. Monitoring data (geophones, ultrasonic devices, and video frames). Chezy coefficient (m 0.5 /s) be accounted for with the AVAL-1D model. The simulated flow cross-sectional areas are generally somewhat lower than the field estimates (Fig. 11), but the agreement appears acceptable for such a simplified modelling approach. The deposition area is reasonably well simulated by the model. About 65% of the total volume is deposited in the retention basin, while the remainder overspilled and flowed further downstream. The simulation shows that the majority of the debris halts in the retention basin, while the tail end of the debris surge remobilizes some of the deposits. This result contradicts the observations from the video recordings. They showed that no remobilization took place, and the de-

12 172 Can. Geotech. J. Vol. 40, 2003 bris remained in the basin (Fig. 6). A more sophisticated modelling approach would be needed to better replicate this flow behaviour during the final stage of the deposition process. Chezy resistance coefficients were estimated from field and monitoring data using eq. [3] (Table 3). These C values are in the range of 4 9m 0.5 /s and are thus on average about 50% smaller than the C values back calculated in the simulations. In the full Voellmy fluid approach used in the simulations, the influence of the dry friction coefficient µ is nonnegligible for the Schipfenbach case, since the two friction terms of eq. [4] are of similar magnitude. This requires a higher C value in the simulations. Another reason for the discrepancy may be the sinuosity of the natural channel ( bouncing of the flow from one bank to the other can be observed on the video pictures) causing additional flow resistance, which is not taken into account in the simulations. Furthermore, the C values back estimated from the direct field observations assume quasi-steady and quasi-uniform flow conditions. Our results may be compared to a number of other debrisflow simulations using the Voellmy fluid approach, where back-calculated model parameters were in the range of C = 11 m 0.5 /s and µ = (Rickenmann and Koch 1997), C = m 0.5 /s and µ = (Ayotte and Hungr 2000), and C =14m 0.5 /s and µ = 0.08 (Jakob et al. 2000). Comparing field observations with computer simulation results involves some uncertainties, and some cautionary remarks should be made. If the simulations are run with a finer reach spacing (e.g., 10 m for the entire flow path), the calculated front velocities show a stronger fluctuation than in Fig. 11. For most of the flow path they vary between 4 and 7 m/s, suggesting that these values may be affected by local bedslope variations (which are not exactly known). Furthermore, there is uncertainty in correlating the exact locations of field cross-sections and those in the simulation. Similar simulation results as presented in Fig. 11 are obtained for front velocities, flow cross-sectional areas, and the deposition zone, if slightly modified pairs of model parameters are selected. An error analysis showed that results calculated by simulations using different C-values (C =9 13 m 0.5 /s) vary from the data illustrated in Fig. 11 by ±17% in velocity and ±1% in flow section. Results obtained using different µ-values (µ = ) vary by ±5% in velocity and flow section. Overall, the comparison indicates that for the Schipfenbach event differences between simulated and observed values are of the order of 20 50%. Although this discrepancy is rather large, we note (in addition to the uncertainties mentioned above) that the simulation of the debris-flow event is based on a number of simplifying assumptions, such as representing the mixture as a quasihomogeneous fluid, introducing constant model parameters along the flow path and within the mass, defining somewhat hypothetical starting conditions, and not including any mass changes by erosion or deposition. In particular, the Chezy resistance coefficient represents both boundary and internal friction of the debris flow, and both channel roughness and fluid properties may change along the flow path as well as within a surge. For similar debris-flow events as in the Schipfenbach, it appears that both a more sophisticated modelling approach and more detailed field data are necessary to possibly improve the agreement between simulated and observed values. The Illbach events Eight debris flows occurred in the Illbach in the year 2000, detailed data were recorded for three events. Magnitude, velocity, and flow depth of these events are listed in Table 2. The volumes of the debris flows were estimated using the hydrographs as obtained by the radar device and thus include the solid as well as the liquid portion of the flows. The values of maximum discharge, maximum flow depth, and velocity were estimated from the video images. Since the videos show the debris flows passing over a check dam with a known width and height, the estimates of the maximum discharge and maximum flow depth are considered reliable. Next, we describe the three debris-flows (3 June, 28 June, and 24 July) and then focus on the largest of the events. The analysis of the data shows significant differences in the flow behaviour of the three debris flows. This variation can be explained by the variability of the solid:liquid ratio, particle size distribution, and woody debris content. The event on 3 June, for example, was characterized by one surge with a high sediment concentration and a low flow velocity (Table 2). In contrast, the debris-flow event on 24 July included two surges: a small initial one with a high water content, woody debris, and a high velocity (4.5 m/s) and a second larger surge with large boulders, no wood, and a velocity of 2.7 m/s. Moreover, five additional debris-flow or hyperconcentrated flow events without complete recordings were observed during Two events occurred in the summer: one on 21 August and another on 20 September. Both of them stopped in the upper part of the fan before reaching the geophones. Three events, presumably a combination of hyperconcentrated flows and debris flows, took place during a regional flood event of October. However, these events occurred during the night and no video images are available because spotlights were not installed in the Illbach. The largest of the debris-flow events observed in 2000 occurred in the early afternoon on 28 June and contained a total volume of m 3. The debris flow was triggered by intense rainfall from a storm passing over the drainage basin. The geophone Geo1 (Fig. 2b) triggered the other geophones, the radar device, and the video camera. Analysis of the video images showed many of the typical debris-flow characteristics, such as a very low discharge just before the first surge, the accumulation of large boulders at the debris-flow front, and a rapid decrease of the flow depth behind the front (Figs. 5d, 5e, and 5f). After the passage of the frontal surge the flow behaviour at the surface appears laminar. The analysis of the video images was compared with the data obtained by the radar device. Figure 12 illustrates the data measured by the radar device and shows the changes of flow depths during a time interval of 1 h. The shape of the debrisflow surge is characterized by a strong and abrupt increase in the flow depth at 14:20 suggesting the arrival of the frontal surge of the debris flow and an exponential decrease of

13 Hürlimann et al. 173 Fig. 12. Radar hydrograph and geophone data of the Illbach debris flow on 28 June flow depth in the wake of the frontal surge. The maximum flow depth is 2.8 m, exceeding the estimate from the video images 100 m upstream by about 0.5 m. This discrepancy can be explained by the differences of the cross sections and the narrow channel at the position of the radar device. In contrast to the data measured by the radar device, the geophone data are more difficult to analyze since the characteristics of the geophone location strongly influence the data. Geophone Geo2 was mounted on a check dam and provided satisfactory signals, whereas the data obtained from Geo3 are diffuse because Geo3 was installed on a sheet pile located on a debris levee. The geophone data are illustrated in Fig. 12 for comparison with the radar device data. Since Geo1 only triggers the measuring devices downstream, the data of Geo2 and Geo3 are provided in Fig. 12. The signals of Geo2 correlate well with the flow depth data obtained from the radar device. The impulses increase significantly at about 14:18 and decrease smoothly afterwards. The data from Geo3 only resolve the front of the debris flow that caused a strong increase in the number of impulses at about 14:19. The final part of the surge was not recorded correctly due to weak input signals. The radar and the geophone data were used to obtain velocity estimates. The different passing times are illustrated in Fig. 12. The mean front velocities obtained for this debris flow are 4.4 m/s (between Geo2 and Geo3) and 3.8 m/s (between Geo3 and the radar device). These calculations and additional data, such as the mean channel bedslope and the flow depth, can be used to estimate resistance coefficients representing both boundary and internal friction of the debris flow, for example, by applying the Chezy equation assuming quasi-steady and quasi-uniform flow conditions to the data given in Table 2. The following C values were estimated from the field observations: C =4m 0.5 /s for the event of 3 June 2000 and C =11m 0.5 /s for the events of 28 June and 24 July The rather low C value determined for the first event can be explained by the small water content of this debris flow. Conclusions Although they provide essential information towards improving the understanding of debris-flow mechanics, realtime data on debris flows in torrent channels are scarce. The following conclusions were drawn from the analysis of field and monitoring data, as well as debris-flow modelling. (1) All debris flows were triggered by intensive rainfalls by in-channel mobilization. The Schipfenbach event initiated in a colluvial channel reach downstream of a steep rocky zone. (2) The mobilized volume of the four monitored debris flows ranged from to m 3. The comparison with historic data showed that events of such magnitude are quite common in both torrents. Large landslides took place in both drainage basins and thus provided a large amount of unconsolidated material in the upper parts of the catchments. (3) The dynamic analysis of the Schipfenbach event shows that the general flow behaviour may be simulated with a simple two-parameter model representing a Voellmy fluid. Such an approach has previously been applied in snow avalanche modelling, and more recently it was tested for the simulation of debris flow events. (4) The analysis of the monitoring data yielded information on the suitability and limitations of the measurement equipment. The ultrasonic and the radar sensors showed that they are practical in defining a debris-flow hydrograph. Moreover, these sensors provide reliable estimates for mean velocity along the selected reach. In contrast, the quality of the geophone data strongly depends on the nature of the contact between the geophone mounting point and the substrate. Channel sections composed of bedrock are ideal locations, whereas unconsolidated materials can provide ambiguous data. Video images can be used to avoid misinterpretations of the monitoring data. (5) The monitoring data indicate a wide spectrum of debrisflow behaviour. Even individual debris flows in the same torrent channel (three events at the Illbach) showed different

14 174 Can. Geotech. J. Vol. 40, 2003 flow behaviours. These differences include variable front velocities (~2 to ~5 m/s), variable flow depths ( m), and variable peak discharges ( m 3 /s). Acknowledgements The debris-flow observation stations were supported by the Board of the Swiss Federal Institutes of Technology, the Swiss Federal Office for Water and Geology, and the Cantons Wallis and Uri. We thank M. Zimmermann and B. Huber for providing historical data, E. Bardou for field data of the Illbach, and B. McArdell for improving the English text. The comments of two anonymous reviewers helped to improve the manuscript. References Arattano, M., and Marchi, L Video-derived velocity distribution along a debris flow surge. Physics and Chemistry of the Earth, 25(9): Arattano, M., and Savage, W.Z Modelling debris flows as kinematic waves. Bulletin of the International Association of Engineering Geology, 49: Arattano, M., Deganutti, A.M., and Marchi, L Debris flow monitoring activities in an instrumented watershed on the Italian Alps. In Proceedings of the 1st International Conference on Debris Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, San Francisco, Calif. Edited by C.-L. Chen. American Society of Civil Engineers (ASCE). pp Ayotte, D., and Hungr, O Calibration of a runout prediction model for debris-flows and avalanches. In Proceedings of the 2nd International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Taipei, Taiwan. Edited by G.F. Wieczorek and N.D. Naeser. A.A. Balkema, Rotterdam. pp Bartelt, P., Salm, B., and Gruber, U Calculating dense-snow avalanche runout using a Voellmy-fluid model with active/passive longitudinal straining. Journal of Glaciology, 45(150): Berti, M., Genevois, R., Simoni, A., and Tecca, P.R Field observations of a debris flow event in the Dolomites. Geomorphology, 29(3 4): Berti, M., Genevois, R., La Husen, R., Simoni, A., and Tecca, P.R Debris flow monitoring in the Acquabona watershed on the Dolomites (Italian Alps). Physics and Chemistry of the Earth, 25(9): Caine, N The rainfall intensity-duration control of shallow landslides and debris flows. Geografiska Annaler, 62A: Choi, S.U., and Garcia, M.H Kinematic wave approximation for debris flow routing. In Proceedings of the 25th International Association of Hydraulic Engineering and Research (IAHR) Congress, Technical Session B, Tokyo, Japan. IAHR, Madrid, Spain. Vol. III, pp Costa, J.E Hydraulic modeling for lahar hazards at Cascades volcanoes. Environmental and Engineering Geosciences, 3(1): GEO SPACE Manual of Geophone GS 20 DX, GEO SPACE, Calgary, Alta. GEO Geomorphologische Analyse des Illgrabens, Bundesamt für Wasser und Geologie (BWG), Biel, Switzerland. Hungr, O A model for the runout analysis of rapid flow slides, debris flows, and avalanches. Canadian Geotechnical Journal, 32(4): Hunt, B Newtonian fluid mechanics treatment of debris flows and avalanches. Journal of Hydraulic Engineering, ASCE, 120(12): Imran, J., Parker, G., Locat, J., and Homa, L D numerical model of muddy subaqueous and subaerial debris flows. Journal of Hydraulic Engineering, 127(11): Iverson, R.M., and Denlinger, R.P Flow of variably fluidized granular masses across three-dimensional terrain 1. Coulomb mixture theory. Journal of Geophysical Research, 106: Iverson, R.M., Denlinger, R.P., LaHusen, R.G., and Logan, M Two-phase debris-flow across 3-D terrain. In Proceedings of the 2nd International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Taipei, Taiwan. Edited by G.F. Wieczorek and N.D. Naeser. A.A. Balkema, Rotterdam. pp Jakob, M., Hungr, O., and Thomson, B Two debris flows with anomalously high magnitude. In Proceedings of the 1st International Conference on Debris Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, San Francisco, Calif. Edited by C.-L. Chen. American Society of Civil Engineers (ASCE), pp Jakob, M., Anderson, D., Fuller, T., Hungr, O., and Ayotte, D An unusually large debris flow at Hummingbird Creek, Mara Lake, British Columbia. Canadian Geotechnical Journal, 37: Johnson, A.M., and Rodine, J.R Debris flow. In Slope stability. Edited by D. Brunsden and D.B. Prior. John Wiley & Sons, New York, N.Y. pp LAMBRECHT Product information Precipitation : Automatic rain gauge LAMBRECHT, Göttingen, Germany, 36 pp. Laigle, D., and Coussot, P Modélisation numérique des écoulements de laves torrentielles. In Colloque d hydrotechnique de la Societe hydrotechnique de France, Session No Edited by CEMAGREF Societé hydrotechnique de France, Paris. MacArthur, R.C., and Schamber, D.R Numerical methods for simulating mudflows. In Proceedings of the 3rd International Symposium on River Sedimentation, Mississippi, pp Nakagawa, H., and Takahashi, T Estimation of a debris flow hydrograph and hazard area. In Proceedings of the 1st International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, San Francisco, Calif. Edited by C.-L. Chen. American Society of Civil Engineers (ASCE). pp Nakagawa, H., Takahashi, T., and Satofuka, Y A debris-flow disaster on the fan of the Harihara River, Japan. In Proceedings of the 2nd International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Taipei, Taiwan. Edited by G.F. Wieczorek and N.D. Naeser. A.A. Balkema, Rotterdam. pp Rickenmann, D Empirical relationships for debris flows. Natural Hazards, 19(1): Rickenmann, D., and Koch, T Comparison of debris flow modeling approaches. In Proceedings of the 1st International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, San Francisco, Calif. Edited by C.L. Chen, American Society of Civil Engineers (ASCE). pp Rickenmann, D., and Weber, D Flow resistance of field and experimental debris flows in torrent channels. In Proceedings of the 2nd International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Taipei, Taiwan.

15 Hürlimann et al. 175 Edited by G.F. Wieczorek and N.D. Naeser. A.A. Balkema, Rotterdam. pp Rickenmann, D., and Zimmermann, M The 1987 debris flows in Switzerland: documentation and analysis. Geomorphology, 8: Sartoris, G., and Bartelt, P Upwinded finite difference schemes for dense snow avalanche modeling. International Journal for Numerical Methods in Fluids, 32: Shieh, C.-L., Jan, C.-D., and Tsai, Y.-F A numerical simulation of debris flow and its application. Natural Hazards, 13(1): Suwa, H Field observation of debris flow. In Proceedings of the Japan China (Taipei) Joint Seminar on Natural Hazard Mitigation, Kyoto, Japan, pp Swiss Federal Office for Water and Geology Hydrological atlas of Switzerland. Federal Office for Water and Geology, Biel. Takahashi, T Debris flow. International Association of Hydraulic Engineering and Research (IAHR) Monograph. A.A. Balkema, Rotterdam, 165 pp. Takahashi, T., Nakagawa, H., Harada, T., and Yamashiki, V Routing debris flows with particle segregation. Journal of Hydraulic Engineering, 118(11): Tognacca, C Murgangentstehung im Gerinne. Mitteilung der Versuchsanstalt für Wasserbau, Hydrologie und Glaziologie, ETH Zürich, Nr. 164, 259 pp. Varnes, D.J Slope movement types and processes. In Landslides: Analysis and control. Edited by R.L. Schuster. National Academy of Sciences, Special Report 176, Washington, D.C. pp VAW Murgänge 1987: Dokumentation und Analyse. Unpublished Report. No. 97.6, Versuchsanstalt für Wasserbau, Hydrologie und Glaziologie (VAW), ETH Zürich, 620 pp. VEGA Product information: Radar. VEGA, Schiltach, Germany, pp VEGA Product information: Ultrasonic level measurement. VEGA, Schiltach, Germany, pp Weir, G.J Kinematic wave theory for Ruapehu lahars. New Zealand Journal of Science, 25: Wilson, R.C., and Wieczorek, G.F Rainfall thresholds for the initiation of debris flows at La Honda, California. Environmental and Engineering Geoscience, 1(1): Zhang, S A comprehensive approach to the observation and prevention of debris flows in China. Natural Hazards, 7: Zimmermann, M., Mani, P., and Romang, H. 1997a. Magnitudefrequency aspects of alpine debris flows. Eclogae Geologicae Helvetiae, 90(3): Zimmermann, M., Mani, P., and Gamma, P. 1997b. Murganggefahr und Klimaänderung ein GIS basierter Ansatz. Vdf Hochschulverlag AG (ETH Zürich), Zürich, 161 pp.