The assessment of rockfall hazard at the base of talus slopes1

S.G. EVANS
Geological Survey of Canada, 601 Booth Street, Ottawa, Ont., Canada KIA OE8
AND

0 . HUNGR Thurber Engineering Ltd., Suite 200-1445 West Georgia Street, Vancouver, B.C., Canada V6G 2T3 Received September 25, 1992 Accepted March 18, 1993
Fragmental rockfall is characterized by the independent movement of individual rock fragments after detachment from a rock face. The continued operation of the process leads to the accumulation of talus slopes. On talus slopes the rockfall shadow extends beyond the base of the talus and consists of scattered boulders that have run out beyond the base of the slope. The landing probability of boulders in the shadow is examined; return periods of the order of 1000 years relative to a house site are typical. Rockfall behaviour particularly with respect to run out into the shadow can be assessed using geological evidence, empirical methods, physical modelling, and computerbased analytical models. An empirical minimum shadow angle of 27.5" (i.e., the angle between the distal limit of the shadow and the top of the talus slope) is suggested and would be useful in rockfall vulnerability studies at the base of talus slopes as a first approximation to shadow limits. It is preferable to the use of the rockfall fahrboschung as proposed by several authors. A random collision lumped mass model (ROCKFALL) is outlined. ROCKFALL uses two restitution coefficients and a transition to rolling criterion. ROCKFALL is used to analyse two fatal rockfall accidents in southern British Columbia, at Hedley in 1939 and Sunnybrae in 1983, which are documented in detail. An additional nonfatal incident is also analysed (Barnhartvale in 1974). Results based on an initial calibration were encouraging. Documentation of the three rockfall incidents shows that, in each case, rockfall fragments impacted on homes at equivalent shadow angles of 30" or more. This would suggest that a review of existing development within rockfall shadow areas at the base of talus slopes may be in order. Key words: rockfall, dynamics, talus slopes, landslides, British Columbia L'Cboulement rocheux fragmentaire est caractCrisC par le mouvement independant de fragments rocheux individuels aprbs dktachement de la face rocheuse. L'action continue de ce processus conduit h l'accumulation de pentes de talus. Sur les pentes de talus, 1'Cboulement rocheux Ctend son empreinte au-del8 du pied du talus consistant en des cailloux CparpillCs qui ont poursuivi leur course au-delh du pied du talus. La probabilitC d'atterrissage de cailloux dans cette empreinte est examinCe; des pCriodes de rCcurrence de l'ordre de 1000 ans par rapport h un site rCsidentiel sont typiques. Le comportement d'Cboulement rocheux, particulibrement par rapport h la distance de parcours dans l'empreinte, peut Ctre CvaluC au moyen de donnCes gCologiques, de mCthodes empiriques, de modClisation physique, et de modkles analytiques pouvant &tretrait& h l'ordinateur. Un angle minimum empirique de l'empreinte de 27,5" (i.e., l'angle entre la limite distale de l'empreinte et le sommet de la pente talus) est suggCrC et serait utile dans les Ctudes de vulnCrabilitC aux Cboulements rocheux au pied des pentes de talus comme premibre approximation des limites des empreintes. I1 est prCfCrable de l'utiliser plutBt que le fahrboschung d'kboulement rocheux tel que proposC par plusieurs auteurs. Un modkle de collision alCatoire de masse de blocks (ROCKFALL) est d6crit. ROCKFALL utilise deux coefficients de restitution et une transition au critkre de roulement. ROCKFALL est utilisk pour analyser deux accidents mortels d7Cboulement rocheux dans le sud de la Colombie-Britannique, h Hedley en 1939 et Sunnybrae en 1983, qui sont docume~tks en dCtail. Un autre incident non mortel est Cgalement analysC (Barnhartvale en 1974). Les rCsultats basks sur la calibration initiale sont encourageants. La documentation des trois incidents d'Cboulement montre que dans chaque cas des fragments d'Cboulement ont frappC les maisons sous des angles Cquivalents d'empreinte de 30" ou plus. Ceci suggCrerait qu'il y aurait lieu de revoir les dCveloppements existants sur les ernpreintes d'Cboulements rocheux au pied des pentes des talus. Mots cle's : Cboulement rocheux, dynamique, pentes de talus, glissements, Colombie-Britannique. [Traduit par la rCdaction]
Can. Geotech. J. 30. 620-636 (1993)

Introduction
Rockfall (Fig. 1) is a slope process involving the detachment of rock fragments and their fall and subsequent bouncing, rolling, sliding, and deposition (Varnes 1978; Hutchinson 1988). A rockfall event may involve the displacement of a single fragment o r several pieces. It may also begin by the detachment of a more o r less coherent block that then disintegrates d u r i n g t h e c o u r s e of m o v e m e n t . Rockfall is a c o m m o n process in mountain terrain w h e r e its continued occurrence forms talus slopes (Rapp 1960n, 1960b; Church ' ~ e o l o ~ i c Survey al of Canada Contribution 42290.
Printed in Canada 1 ImprinlC ou Canada

e t al. 1979), a c c u n ~ u l a t i o n sof rock fragments at t h e base of a source rock slope. T h i s paper is concerned'with assessing t h e hazard presented b y events at the lower e n d of the magnitude spectrum, the "fragmental rockfall" of Hungr and Evans (1988, 1989), near the base of talus slopes. Fragmental rockfall i s characterized by a m o r e o r less independent movement of individual particles, as opposed to the sliding o r mass flow o f c o h e r e n t o r b r o k e n r o c k typical o f a r o c k avalanche. I n general, f r a g m e n t a l rockfall involves relatively s m a l l detachments ( < l o 5 m3), although there is n o well-defined volume limit.

EVANS AND H U N G R

FIG. 1. Rockfall from Bastion Mountain, British Columbia on December 23. 1959. Shuswap Lake in foreground (photograph courtesy of Dorothy Brook).

FIG.2. Example of large boulder runout in the Sirnilkameen Valley, British Columbia. Large boulder sits on river terrace beyond the foot of the talus slope and originated in a rockfall from the slopes to the left. Note person for scale. Low-magnitude fragmental rockfall is distinguished from larger rockfalls (>lo6 m3). These invariably disintegrate and are transformed into a rapid movement of ~lnsortedfragments termed a stiirzstrom (Heim 1932), rock avalanche (McConnell and Brock 1904), or rockfall avalanche (Varnes 1978) where inclusion of the term rockfall implies an episode of free fall during the movement. Hungr and Evans (1989) compiled rockfall accident statistics for Canada and found that rockfall has caused only four deaths in houses and nine deaths on transportation routes in this century. Even allowing for the fact that the records are incomplete, this number is surprisingly small. It may be compared with casualties in the mountainous regions of Europe where in Italy in 1969, for example, a rockfall from Monte San Martino caused eight deaths in the town of Lecco (Eisbacher and Clague 1984). This paper enlarges on previous work (Hungr and Evans 1988, 1989), which gave partial results. It addresses the problem of assessing rockfall hazard on and beyond the base of talus slopes. Periodically a rockfall occurs in which one or more larger boulders bounce and roll down the talus slope and move beyond its margin, coming to rest at some

E
g0.8

TALUS SLOPE PROFILE

2
W

2 ,,,

0.4

$ 0.2
0 0 100 200 HORIZONTAL DISTANCE (rn) 300

FIG.3. Typical downslope distribution of mean grain sizes in a talus deposit (data from Worobey 1972). distance from the slope base (Fig. 2). It is these b o ~ ~ l d e r s that c o n s t i t ~ ~ the t e greatest hazard in the vicinity of talus slopes, and uncertainty about their behaviour poses a major difficulty in rockfall-vulnerability assessment in mountain-

CAN. GEOTECI-1. J. VOL. 30, 1993

FIG.4. Schematic diagram of the characteristic rockfall path profile. A-C is the talus slope, with mean angle PI, and C-D is the rockfall shadow, with a shadow angle PZ P3, substrate angle.
X
W

<
w
3 . . . I 4:

DATA FROM WOROBEY (1972)

t o 000

DATA FROM FIG 6

1000-

-10.~

loo10-

-lo2
-10.'

atively rare, the indirect costs of sterilization of land endangered by rockfall are in the order of tens of millions of dollars. Continuing development pressure in the mountainous areas of Canada will necessitate more rigorous geotechnical analysis of potentially hazardous sites. The nature of rockfall deposition on and beyond talus slopes is investigated. Several methods of rockfall analysis are examined and a numerical analytical model (ROCKFALL) is outlined. Three case histories of rockfall accidents at the base of talus slopes in the mountains of British Columbia are described and are analysed with ROCKFALL.

h
LANDING PROBABILITY IMPACT PROBABILITY
W

8 8

2
(0

100

200

300

400

500

HORIZONTAL DISTANCE FROM APEX (m)

The behaviour of rockfall fragments on talus slopes Depositiorz on talus slopes The depositional pattern exhibited in talus slopes reflects certain regularities. An important aspect is the natural "gravity" sorting by size observed by many authors (e.g,. Rapp 19600, 1960b; Gardner 1970; Kotarba and Stromquist 1984; Brundsen et al. 1984; Statham and Francis 1986) wherein the coarsest boulders are found at the base of the slope. Size sorting contributes an important roughness component to the talkis slope surface, which becomes the characteristic path profile for successive rockfalls. A typical example of size sorting on talus slopes is shown in Fig. 3, derived from data collected on a talus cone in the Similkameen Valley, southern British Columbia by Worobey (1972). The mean particle size is seen in this instance to increase gradually from about 5 to 10 c m over approximately 300 In of slope length. Approximately the lowermost 40 m of the cone surface is built up of boulders with characteristic sizes of 1 m or greater. The slope angle reduces to as little as 15O in the coarse basal accumulation. The relatively low variation of grain size in the upper part of the cone is probably the result of redeposition by surficial talus flows, which are common in the area. Roclifrrll path profile A rockfall-dominated talus slope exhibits a typical profile as seen in Fig. 4. Finer talus fragments accumulate below the apex (point A) at an angle of approximately 38". Lower down, the talus angle ranges from approximately 32 to 38O. The lowermost part of the talus deposit contains the largest fragment sizes, and the tallis sl-ope angle sometimes falls

FIG.5. Landing- and impact-probability distributions calculated for the talus slope shown in Fig. 3. ous regions. This has emerged as an important problem in planning resource development in the vicinity of rock slopes in Canada (e.g., Ballivy et al. 1984; Boyd 1991) and is the primary focus of this paper, since, as noted above, although rockfall damage to buildings at the base of talus slopes is rel-

E V A N S AND HUNGR
TOE OF TALUS

623

I
I

END OF SHADOW
APPROX 500*

\
IMPACT SEGEMENT (1 m w~de)

/ & A
POINTS

/ \

/p%2AL MEASURED

FIG.7. Unfolded plan of the cone surface, showing influence area used in calculations of impact frequency.

TALUS FLOW

TOE OF TALUS

END OF SHADOW

I , ~ l l ~ f , , , , , , , 100 200 300 400 HORIZONTAL DISTANCE FROM APEX (m)

500

20

40

60

80

100

120

FIG. 8. Approximate return periods for impact damage to a 10 x 10 m house site. ability of a unit volume of rock fragments landing at any particular location on the cone surface must then be approximately constant. In the Similkameen Valley of British Columbia, talus cones (exposures of the Mazama Ash deposited 6600 years BP; Clague 1980) can be found at a typical vertical depth of 3 m (Worobey 1972). T h e corresponding mean vertical g r o w t h (V) r a t e of t h e talus s u r f a c e is 316600 = 0.00045 mlyear. The rate of growth in the direction normal to the c o n e surface is approximately (cos 34") V, i.e., 0.00037 mtyear (cf. Luckman 1988). The probability of fragments landing on a given square metre plot of the surface is a function of the mean particle diameter:

DISTANCE FROM TALUS TOE (m)

FIG.6. Distribution of boulders in the distal part of a rockfall shadow area surveyed in the Similkameen Valley, British Columbia. Total number of boulders = 571; width of area surveyed = 200 m; mean boulder diameter = 1.8 m. :",based on 50% area coverage. to 10 or 20, as shown between points B and C. Point C is the base of the talus deposit. Beyond this point the slope is no longer completely covered by talus fragments. The average talus angle is P, (Fig. 4). The surface to the right of point C (Fig. 4 ) is termed the substrate surface, consisting of material and landforms predating the talus deposits. That part of the substrate surface covered discontinuously by scattered large boulders that have rolled or bounced beyond the base of the talus (C-D in Fig. 4) is referred to here as the rockfall shadow. The "shadow angle" is defined as the angle between the outer margin of the shadow and the apex of the talus slope (P2 in Fig. 4). The distal part of the shadow often contains only very few boulders, which are sparsely distributed on the surface.

Spatial distribution of rockfall impact on talus slopes Talus cone slope angles remain fairly constant during their gradual formation. As a result, it can be assumed that a relatively uniform thickness of material is added to the entire cone surface over any long period of time. The prob-

is the frequency of landings per year per square where f,, metre of cone surface, R, is the surface growth rate in the normal direction in metres per year and D, is the characteristic dimension of the mean rock fragment, defined as the side of a cube with the same volume as that occupied by the fragment in the deposit. The mean grain size of talus deposits increases with distance from the apex. Consequently, by [ L ] , the frequency of landing decreases. The frequency variation shown by the

624

CAN.

GEOTECH. J . VOL. 30,

1093

f r o m t h e talus t o e , measured at a n o t h e r s i t e in t h e Similkameen Valley. The landing probability in thgshaddk area was estimated by surveying the distribution of boulders in strip-like segments concentric on a talus cone. The measurements were made on a plot of land 200 m wide, extending from the toe of the talus cone to the distal margin of the shadow. The distribution of boulder frequency versus distance from the toe is shown in the form of a histogram with a spacing of 20 m (Fig. 6). The boulders lie on a river terrace with an estimated age of about 5000 years. (J.M. Ryder, personal communication, 1987). The landing probabilities are calculated by dividing the boulder landing counts given in Fig. 6 by 5000 and by the survey area of 20 x 200 m. The resulting values are shown as open triangles in Fig. 5. The two segments of the landing probability curve show continuity of slope, although they originate from two different sites. In Fig. 5 the landing probability estimates in the steep central part of the talus slope are severely distorted by the frequent occurrence of talus flow movements. The probabilities estimated near and beyond the margin are, on the other hand, representative of individual particle deposition. Impact hazard affecting a structure situated on or near the talus cone derives not only from fragments landing at that location but also from those crossing the location and destined to land farther down the slope. This impact frequency, A,,, can be estimated by integrating numerically the landing probabilities over an influence area sketched in Fig. 7:

where i is a counter of sampling points regularly spaced at a slope interval, d is the downslope from the point in consideration, and r is defined in Fig. 7. The point considered is located halfway between a pair of sampling points, as shown in Fig. 7. The calculation results are given as the solid line in Fig. 5 in terms of events per metre width. Return period T of impact events at various locations can be estimated as

FIG.9. Rockfall hazard at Silverhope, British Columbia. (A) Oblique view of Silverhope, showing source cliffs and vegetated talus slopes. Borrow pit referred to in text is cleared area near the ccntre of the photograph. (B) Large boulder in rockfall shadow at Silverhope. broken line with open circles in Fig. 5 has been derived by applying [ I ] to the average grain size profile reported by Worobey (1972) in Fig. 3. The landing probability curve in Fig. 5, constructed on the premise of [I], extends only to the toe of the talus slope. To extend it beyond the toe into the rockfall shadow area, additionai information is required. This is provided by Fig. 6, showing the distribution of rockfall fragments with distance

where W is the width of the area considered. The resulting return periods for a 10-m-wide house site are indicated in Fig. 8. At 50 m from the talus margin, the impact probability is 0.000 05 eventslm, and the corresponding return period for a 10-m house site is 1900 years. A housing subdivision located at the same distance and covering an area 200 m wide would face a probability of being struck by a single boulder of 1 in 95 years. Short-term probabilities can vary considerably from the long-term average. There is some evidence that rockfall rates have decreased throughout the Holocene (e.g., Worobey 1972). In this case, the estimated return periods represent a lower limit estimate.

Prediction of rockfall behaviour. I. Geological evidence In some cases, it is possible to ascertain the past behaviour of a rock slope over a long period on the basis of geologi-

EVANS AND HUNGR

FIG. 10. Profiles of 16 surveyed rockfall paths from British Columbia (modified from Hungr and Evans 1988). The profiles have been plotted with the talus apex as a common point. A, base of the talus slope; 0, distal margin of the rockfall shadow; 0, either known source areas or the crest of the source cliff. 1, Hedley; 2, Similkameen A (Sweatlodge)*; 3, Similkameen B (rockslide); 4-Similkameen C (kame); 5-Similkameen D (Winters Creek); 6-Similkameen E (campground); 7, Similkameen F (speedway); 8, Pukaist; 9, Sunnybrae; 10, Barnhartvale:':; 11, Silverhope A; 12, Silverhope B; 13, Silverhope D; 14, Silverhope C; 15, Hope North*; 16, Stawamus Chief2' (asterisks denote sites where fresh debris was found in the rockfall shadow area).
TABLE 1. ROCKFALL: preliminary model parameters Large rocks on talus (1000 - 20 000 kg) Normal restitution coefficient (k,) (no crushing)" Tangential restitution coefficient (kJh Yield contact force F, (kN)' Initial contact stiffness C, (kN/m)' Stiffness reduction ratio Rc Rolling friction coefficient (h)
"V' = V,k, where VA, is normal velocity component after impact. = V,k, where V:, is tangential velocity component after impact. 'For definition see Falcetta (1985).

Small rocks on clean rock (20-1 00 kg) 0.9 0.9 10 100 0.05 0.52

0.7 20.8 50 100 0.05 0.52

"$

cal evidence. Such information can then be extrapolated s o as to predict the reach of future rockfall within a given return period. An example is provided by the evaluation of rockfall hazard at the community of Silverhope (Fig. 9A) in southwestern British Columbia. The community is located on an alluvial fan in front of a deltaic terrace. Immediately behind the terrace, cliffs composed of massive Eocene quartz diorite and hornfels rise to a height of 600 m (Fig. 9). Extensive talus deposits, and numerous large boulders that have rolled

beyond the talus margins, testify to rockfall activity, although this activity appears to predate the development of the extensive forest cover (Fig. 9). No rockfall accidents have been reported since the site was settled in the 1950s. The rockfall shadow at Silverhope is well defined (see profiles 11-14 in Fig. 10). The terrace surface beyond the rockfall shadow is underlain by a deposit of cross-bedded sand containing no gravel or boulders, which is exposed in a small borrow pit (Fig. 9). The sand is of deltaic origin, deposited in standing water. However, no lake has existed at

626

CAN.

GEOTECH. J. VOL. 30,

1993

FIG. 11. Simulation of full-scale experiments carried out by Broili (1977). (A) Typical ROCKFALL result, with a fragment volume of 0.5 m3. Velocity is shown in upper frame. (B) Comparison of Broili's results and calculations using ROCKFALL, showing the downslope distance d to the point of onset of rolling as a function of fragment volume V. the site in the Holocene. Therefore, the surface of the deltaic terrace must be at least 11 000 years old, dating back to the last stages of glacial ice retreat from the valley (J.J. Clague, personal communication, 1986). The area immediately outside the well-defined rockfall shadow therefore has a mobability of less than 1 : 11 000 of being reached by rockfall from the cliff in a given year. The general frequency of rockfall in this area was indicated by a detailed inspection of talus deposits beneath cliff fronts totalling approximately 2700 m in length. The talus deposits are partly forested and heavily moss covered (Fig. 9). Only six rock fragments were found whose appearance indicated that they have been deposited within the last year, their sizes varying from 0.01 to 0.1 m3. The average frequency of fragments in this size range falling onto the talus apron therefore works out to 0.04 per year per 100 m length of cliffs, which is within the range of the frequency-magnitude curve described by Gardner (1980). No fresh fragments or tree damage associated with rockfall were identified within the rockfall shadow area. Not all sites provide sufficient evidence to allow such reasoning. If it were not possible to reliably date the substrate surface, if weathering conditions change (e.g., because of deforestation or a recent major instability), or if there were boulders of glacial or debris-flow origin impossible to distinguish from rockfall deposits, then this method may not provide satisfactory answers.

Prediction of rockfall behaviour. 11. Empirical evaluation Two related empirical measures have been suggested to estimate the run-out distance of rockfall fragments at the base of talus slopes, viz., the rockfall fahrboschung and the minimum shadow angle.
Rockfall fahrboschung The equivalent of a rockfall fahrboschung has been suggested by Onofri and Candian (1979) in their analysis of 98 rockfalls triggered by the 1976 Friuli earthquake in Italy. The rockfall fahrboschung (P) is the angle between the highest point of the rockfall source scar and the stopping point of the longest run-out boulder for any given rockfall. Onofri and Candian (1979) found the limits of P to be 28.34 and 40.73". They found that the values of P had a Gaussian distribution and were able to assign probabilities to the various values of p. Toppe (1987) uses the same parameter to estimate rockfall runout. Although he does not specify the nature of his field data, he reports that 50% of rockfall fragments stop within p = 45" and 95% stop within P = 32". These values contrast strongly with those of Onofri and

EVANS AND HUNGR

Candian (1979) who found that 50% stopped within P = 33.5" and 72% stopped within P = 32". The calculation of the rockfall fahrboschung requires the measurement of the starting and end points for individual rockfalls.

Minirnclm shadow angle We suggest an alternative approach, following Lied (1977), where an empirical run-out angle, the minimum shadow angle (P2), defined in Fig. 4, is keyed to the apex of the talus slope and not to the source area of the rockfall on the rock slope above. The approach does not necessitate the start and end point of each rockfall to be located, since the effect of rockfall activity is integrated over time by considering the longest boulder runout in a given rockfall shadow. A minimum value for (P2) has been suggested by several authors. Lied (1977) proposed that it should be in the range of 28-30", and Hestnes (1980) suggested a value of 25", although neither author discussed the bases of these estimates. In this work, 16 profiles of rockfall paths from southwestern British Columbia have been compiled (Fig. 10). The profiles have been plotted so as to join at the talus apex as a common point. The points, representing the limit of the shadow, plot consistently just below a line inclined at 27.5", irrespective of the height of the source cliff, length of path, or the substrate angle, and represent a minimum shadow angle. The slopes in Fig. 10 are from locations both in the sparsely forested arid interior of British Columbia and in the humid southern Coast Mountains where there is dense forest cover outside the most active talus areas. The use of an empirical minimum shadow angle of 27.5" would appear to be a useful method for the preliminary estimation of maximum rockfall reach, since the authors have observed that the same angle applies at many other sites in British Columbia. Flatter angles sometimes result where the talus and (or) the substrate surface is relatively smooth. For example, a shadow angle of 24" has been measured by the authors for a single 4-m-diameter boulder rolling or sliding over the smooth surface of a glacier near Whistler, B.C. Rapp (1960b, p. 98) documents a rockfall in which a boulder ran out to a shadow angle of 23" on a surface partly covered by grass and snow. Some rockfall profiles from the Italian Alps (Govi 1977; Govi and Sorzana 1977; Fenti et al. 1977; Paronuzzi 1988, 1989) also suggest a flatter minimum shadow angle may result where the talus surface is relatively smooth. In addition, the impact of large rockfalls on loose talus may form flow-like slides that may run out beyond the rockfall shadow limit. Kolderup (1955) describes an example from Norway in which the flow slide generated by such a process ran out to an equivalent shadow angle of 22" and destroyed two farmhouses. Itzterpretatiotz In our investigations we have observed that long rockfall trajectories on moderate slopes tend to be dominated by rolling. This is in agreement with full-scale experimental observations made previously by such workers as Ritchie (1963), Broili (1977), and Kirkby and Statham (1975). This observation helps to explain the concept of a limiting shadow angle discussed above. T h e kinetic energy acquired by the fragment in the initial fall is largely lost in the first impacts on the talus surface near the top of the

FIG. 12. Map of Hedley, British Columbia, showing location of 1939 rockfall release point and most probable impact zone. Contours in feet.

slope. Thus, regardless of the height of fall, each fragment begins rolling near the top of the slope and its final runout into the shadow is approximated by projecting the slope of the energy line, i.e., the rolling friction gradient, from the top of the talus slope. The value of P2 would thus appear to be the lower limit of the rolling friction angle for large boulders rolling over finer talus, and its variation would be expected to reflect differences in the surface characteristics of both talus slope and substrate. The limited horizontal reach of boulders from Stawamus Chief (profile 16, Fig. 10) confirms that the initial momentum from fragments falling from a near-vertical source is practically lost in the first series of impacts. Long horizontal reach is achieved only where fragments are able to build up sufficient horizontal and rotational momentum by travelling over a long inclined segment of talus slope. The rockfall fahrboschung ignores the marked assymetry in energy expenditure noted above and is dependent on the height of fall. In assessing rockfall vulnerability at the base of a talus slope, use of the fahrboschung to delimit run-out distances requires knowledge about future detachment locations on the rock slope above. A conservative approach in these circumstances may involve taking the minimum fahrboschung angle from the top of the rock slope (Toppe 1987) which would overestimate the runout at a site.

Prediction of rockfall behaviour. 111. Development of an analytical model In recent years, many investigators turned their attention towards computer-based analytical .models, simulating the

C A N . GEOTECH. J . VOL. 30, 1993

FIG. 13. Area of Hedley affected by 1939 rockfall. Photograph taken in summer of 1939. Note location of houses close to the base of the talus slope within the rockfall shadow area (photograph courtesy of Helen Moore, Hedley Heritage Arts and Crafts Society).

actual progress of a rock fragment down its path (Piteau and Clayton 1977; Azimi and Desvarreux 1977; Azimi et al. 1982; A s t t et al. 1984, 1987; Wu 1985; Falcetta 1985; Bozzolo and Pamini 1986; Rochet, 1987~1,1987b; Kobayashi and Kagawa 1987; Descoeudres and Zimmerrnan 1987; Spang 1987; Spang and Rautenstrauch 1988; Bozzolo et al. 1988; Paronuzzi 1988, 1989; Pfeiffer and Bowen 1989; Kobayashi et al. 1990; Fornaro et al. 1990; Azzoni et al. 1992). Hungr and Evans (1988, 1989) developed a random collision model (ROCKFALL) which is outlined below. Bo~lrzcingmode ROCKFALL is based on a lumped mass model with separate normal (k,) and tangential (kt) restitution coefficients (Azimi and Desvarreux 1977; Azimi et al. 1982; Wu 1985), one of which (k,) is momentum scaled to allow for plastic deformation during impact. The assumption of constant restitution coefficients is reasonable only in cases where each impact involves approximately the same level of momentum normal to the contact surface. This is obviously not the case on paths where, for example, the first impact on the talus surface follows a long free flight from the high cliff. According to full-scale experiments by Broili (1974), 75-86% of the kinetic energy gained in the initial fall is lost in the first impact. Subsequent impacts, involving lesser normal momenta, consume much less energy. It is for this reason that the bilinear contact deformation model developed by Falcetta (1985) was incorporated in ROCKFALL (Hungr and Evans 1989). Preliminary calibration of the model for large rocks was achieved using the results of Broili (1977). Figure 11 shows a simulation of one of the experiments reported by Broili (1977) in which a camera recorded three modes of motion: (i) the initial fall from a height of 210 m, (ii) a sequence of bounces, and (iii) a long segment dominated by rolling movement. The three modes are simulated by ROCKFALL in Fig. 11A. According to Broili's observations, as reproduced in Fig. 11B, the length of the zone of bouncing increased
A

with decreasing volurne of the rock fragment released. This result is plotted in Fig. 11B over a range of 0.5-10 m3. By trial and error iterations of ROCKFALL, we have been able to approximate the experimental relationship as shown by the broken line in Fig. l l B , by using ROCKFALL parameter values given in the first column of Table 1. For preliminary purposes only, in the absence of more detailed calibration, these parameters are considered applicable to impact of large rocks (1000 - 20 000 kg mass) on talus surfaces, without heavy vegetation. It is unlikely that the same parameters would apply for smaller rocks bouncing on different surfaces. For small rocks contacting clean rock surfaces, we have back analysed the results of experiments carried out by Wu (1985). The model parameters providing a reasonably good fit to these r e s ~ ~ lare t s given in the second column of Table I. It must be stressed that more detailed calibration based on controlled experiments is necessary. Transition to r-olliizg inode The bouncing-only model does not simulate correctly that stage of rockfall movement when bounces become negligibly short and rolling begins. It is therefore necessary to provide a transition at some point into a rolling mode, simulated by a frictional model. Several different criteria for transition between the bouncing and rolling mode were tried, including a limiting length of bounce, limiting time, and velocity. With all of these, there were problems with numerical instability of the model, as they allowed bounces o'f extremely short length or low amplitude. The most satisfactory results were finally reached with our transition to rolling criterion described in the following paragraphs. ROCKFALL calculates the "energy head" of the fragment at every point during its descent, as indicated by the broken line in Fig. 11B. While in trajectory, the fragment has a constant energy "head", E:

EVANS AND HUNGR

FIG. 14. Archive photographs of damage caused by 1939 Hedley rockfall. (A) Boulder which killed the two victims of the accident in the house behind. (B) Damage to a room at the corner of the Moore residence where the two Moore children escaped injury (photographs courtesy of Helen Moore, Hedley Heritage Arts and Crafts Society).

where V is the length of the velocity vector, z is the fragment elevation, and g is acceleration due to gravity. After an impact, the velocity component normal to the path is reduced by a ratio k , and the tangential velocity component by kt. It can be shown that the resulting incremental loss of energy (AE) is k: +k,2 tan2 0

1 + tan' 0
where 0 is the angle of incidence prior to the impact. Thus, the kinetic energy is reduced in each impact by a ratio ranging from k: for very flat trajectories, through (k: + kf)/2 for 45" impacts, to k i for steep trajectories approaching the perpendicular. The "energy line" resulting from plotting the relationship in [4] appears as a series of steps, separated by horizontal

lines (Fig. l l A ) , the length of which equals the trajectory length AL. When the ratio AEIAL is less than the tangent of the slope angle, the fragment accelerates continuously and the energy line rises above the path. When the ratio becomes greater than the slope gradient, the fragment decelerates and the trajectories rapidly become shorter. Deposition of the fragment then occurs. In ROCKFALL, a transition into rolling mode is made when, during three consecutive bounces, the ratio between the kinetic energy head lost in an impact and the horizontal length of the corresponding bounce (AEIAL) is greater than the rolling-friction coefficient, i.e., the rolling movement mode becomes more efficient than the bouncing mode. When there is a change in path slope, such as at the crest of a small cliff, the fragment may temporarily move back into trajectory. For long slope segments that have slope angles 545", ROCKFALL generally predicts rockfall fragment paths that

630

C A N . GEOTECH. J. VOL. 30, 1993

1000

------ Topography

Impact on houses

I
0 200 400 600 800 1000

metres

FIG. 15. The 1939 Hedley rockfall. ROCKFALL silnulation of the longest rockfall path with k,, (normal restitution coefficient) = 0.70, k , (tangential restitution coefficient) = 0.80, and rolling friction coefficient = 0.52. are dominated by rolling movement. This is exemplified by the long travel paths sim~ilatedin the Hedley and Sunnybrae examples below, and it is in agreement with the empirical observations of Ritchie (1963). Under such circumstances, the variation and even the absolute value of the restitution coefficients are of secondary importance compared with the magnitude of the chosen rolling-friction coefficient.

FIG. 16. Oblique aerial view of Sunnybrae (S), British Columbia. Note the source cliffs for rockfall from Bastion Mountain. E, location of release point of 1983 rockfall; @, location of damaged house in which two people were killed. One side of the wedge formed along a faulted contact plane between basalt in the footwall and recrystallized limestone in the hanging wall, dipping at a mean angle of 52", a few degrees to the left of the sliding direction. The opposite side of the wedge consists of a series of near-vertical cross joints striking perpendicular to the fault plane. Light-coloured, relatively fresh rock at the upper end of the scar, some 50 m above the talus, indicates that the volume of the detached rock was about 5000 m3. A few large limestone boulders lie on the talus surface about 100 m below the apex. These boulders have a median size of approxilnately 0.7 m and a range of 1.5 m and are partly buried by the fine basaltic talus. The limestone rock that formed the rockfall is laminated b~it strong and sparsely jointed, tending to break into large blocks. In contrast, the underlying basalt is closely fractured. No traces of the rockfall path were found on the talus, except for the few remnant boulders mentioned above. The 1939 accident occurred in a newly built part of town known as the Ready Cash subdivision, consisting of two streets parallel to the toe of the talus cone (Figs. 12 and 13). Several boulders reached the subdivision, as reconstructed by arrows in Fig. 12. The boulders ranged from 1 to 2 m in size. Four houses were severely damaged (Fig. 14). A number of persons were awakened from their sleep by the walls of their bedrooms crushing under impact (Fig. 14). Fortunately, as noted above only two deaths occurred. 'The rockfall impact in the subdivision took place within the rockfall shadow at an equivalent shadow angle of 30". The rockfall fahrboschung for the Hedley event is 34". Meteorological d a k from Hedley were analyzed for the 1939 event but do not suggest an obvious meteorological trigger for the rockfall. T h e discontinuity patterns in t h e southeast face of Stemwinder Mountain indicate that similar conditions of detachment exist for fiiture rockslope failures along the face. Although homes were moved from the run-out area of the 1939 event (B.C. Order in Council 1576), several homes at

Case histories of rockfall accidents at the base of talus slopes Three case histories of rockfall accidents at the base of talus slopes in British Columbia were investigated in detail. The sites were visited in September 1987 and data collected on the geology of the detachment zone and the geometry of the path. Detailed theodolite surveys were conducted on the path and eyewitnesses were interviewed in the first two cases discussed.
Hedley, B.C., Jatzuary 17, 1939 Hedley is a small mining c o m m ~ i n i t ylocated in t h e Sirnilkameen Valley in the interior of British Columbia. Part of the community was situated directly beneath the rock faces and talus slopes of Stemwinder Mountain (Fig. 12). A rockfall from Stemwinder Mountain struck the town at 01:OO on Burns Night, January 17, 1939, killing two people and damaging several houses. The slopes above the townsite are ~inderlainby the Upper Triassic Nicola Group, consisting of an interbedded, folded, and sheared sequence of basalts and limestone. The source of the 1939 rockfall is above the apex of a large talus cone consisting of relatively uniformly graded basalt fragments. Most of the talus surface consists of cubical fragments with median sizes of approximately 4 0 mm, ranging from 10 to 150 mm. The surface is susceptible to shallow tallis sliding and is free of vegetation. Its profile is uniform at 35". The source area of the rockfall has been identified with the help of eyewitnesses. It is a large asymmetric wedge scar.

EVANS A N D HUNGR

FIG. 17. 'The 1983 Sunnybrae rockfall. (A) The damaged house in which two people were killed (photograph courtesy of Salmon Arm Observer). (B) 'The boulder involved in the rockfall being buried where it came to rest.

the present Hedley townsite are located beneath rockslopes and talus that are within the zone defined by the empirical minimum shadow angle of 27.5". The profile of the longest rockfall path is shown in Fig. 15. The path was simulated with ROCKFALL, and the calculated reach distance agrees well with the actual experience. The results of the computer simulation again confirm the predominant rolling mode of movement. It is noted that the analysis could for practical purposes be replaced by one based on a frictional model. Sunnybrae, B.C., Noveinber 23, 1983 Sunnybrae is a small community in the British Columbia interior located on the north shore of Shuswap Lake (Fig. 16) across from the town of Salmon Arm. The community lies at 350 m above seal level (asl) directly beneath the precipitous faces of Bastion Mountain (Fig. 16), which extends to 1300 m asl. The cliffs are made LIP of metamorphosed limestone of the Sicamous Formation. The steep, sparsely forested slopes extending from the foot of the cliffs down towards the lakefront are underlain by

weak schistose bedrock units, thinly mantled by fine-grained talus derived from the same rock. The bedrock is exposed in numerous small cliffs and outcrops on the slope. Only very few boulders can be observed in the wooded rockfall shadow area surrounding Sunnybrae. It appears that the main product of weathering of the slope above is a relatively fine grained talus material. Comparison of photographs of the precipitous faces of Bastion Mountain taken in 1905 and 1983 by Geological Survey of Canada field personnel indicate little change in the morphology of the faces and therefore limited rockfall activity in that time interval. In 1959, however, a substantial rockfall from the main face of Bastion Mountain (Fig. 1) occurred, suggesting that rockfall is a continuing, if infrequent, process in the vicinity of Sunnybrae. At 01:50 on November 23, 1983, a wheel-shaped boulder of 6 x 6 x 2 m detached near the base of the limestone cliffs, rolled, and bounced down a path 790 m long. It partly destroyed a house at the north end of Sunnybrae and killed two inhabitants (Fig. 17). The boulder demolished approximately one half of the house, leaving the other half rela-

CAN. GEOTECH. I . VOL. 30, 1993

'"""

PROFILE OF PATH

300

I 0

100

I 200

I I 300 400 HORIZONTAL DISTANCE

I 500

I 600

I
I 700

(m)

--I

PLAN OF PATH
Source

" . " . , . , . ", " " '* "us-

Lz
%Impact

ZP~'"':'

''

(Post

- Event)

House M a r k s Reported b y Miller ( 1 9 8 3 )

FIG.18. Path of the 1983 Sunnybrae rockfall in profile and plan surveyed in 1987.

tively intact. A person sleeping in that part of the house escaped injury. A second house was missed literally by a few centimetres (Miller 1983). The release point of the boulder is shown in Fig. 16. It is in a vertical part of the cliff face some 25 m above its base. The rock is banded crystalline limestone, light to dark grey in color and very strong. The boulder separated along an exfoliation joint dipping 70" to the south (down the rockfall path). The upper edge of the fragment separated from the rock above along a curving near-horizontal joint, leaving an arch-like overhang over 2 m deep. The exfoliation joint can be observed to continue upwards into the overhang. It appears partly open, and slight groundwater flow issues from it, staining the face below (the rockfall boulder was reported by Miller (1983) to have had brown staining on one face). The eastern lateral boundary of the block was defined by a joint dipping approximately 50" to the west. It is clear that the detachment of the boulder resulted in the creation of an overhang more exposed than that which existed before. The possibility cannot be ruled out that the event of November 1983 was initial to a series of detachments from the same source area and with similar mechanics.

The travel path of the boulder was reconstructed by surveying the marks left along its path, most of which were still discernible in 1987 (Fig. 18). The marks were of two types: ( i ) impact craters, typically 4 x 1 m in plan and 0.3-0.5 m deep; and (ii) elongated rolling marks, resembling a trench 2 m wide and 0.5 m deep, up to 30 m long. As shown in the profile in Fig. 18, the boulder moved by rolling throughout most of its descent. A long bounce occurred below the crest of the lower cliff, followed by several shorter bounces and abrupt changes in direction, due to topographic detail (see plan in Fig. 18). Small fragments broke off from the boulder at several places near the end of its path. These fragments were generally found far from the impact craters, indicating that they separate'd during rotation of the boulder. A simulation of the rockfall path using ROCKFALL appears in Fig. 19. The boulder mass was assumed to be 150 t, so most impacts were plastic. It is of interest to note that the length of the parabolic trajectory at the lower cliff in the simulation is very close to the length observed. Therefore, the velocity estimate shown in the upper part of the figure appears realistic. The analysis indicates that the relative shortness of the path, characterized by an equivalent shadow angle of 34" and a rockfall fahrboschung of 35", is due to the

EVANS A N D H U N G R

633

plastic energy losses incurred in bouncing below the lower cliff (Fig. 19). Had it continued rolling, the boulder may have travelled farther. It is of concern that many homes are located beneath these cliffs within the zone defined by the empirical minimum shadow angle of 27.5" discussed above. Bnmhnrtvnle, spring 1974 The Barnhartvale subdivision is located within the City of Kamloops in the interior of British Columbia. The western part of the community is located at the foot of some prominent east-facing rock slopes and associated talus (Fig. 20). The rock is intensely jointed metavolcanics of the Cache Creek Group, with prominent joint sets dipping at about 50" into and out of the slope. These discontinuities form numerous detachment surfaces for fragmental rockfall, a fact which is reflected in the substantial talus accumulation downslope. We have been able to reconstruct t h e event ~ i s i n ga 1:2500 orthophoto with 10-m contours and our 1987 field observations. In the spring of 1974, a boulder measuring about a metre 775 m as1 on the rock slope, in diameter, fell from a b o ~ i t bounced, and rolled over the talus slope and beyond its base. The rock severely damaged a house on Lot 16, Ronde Lane at 5 9 2 m asl. Fortunately no one was hurt. Impact took place at an equivalent shadow angle of 30". The rockfall fahrboschung is 34". The incident led to the implementation of the following mitigative measures: scaling of the rock slopes, which included the removal of problematical rock masses by blasting; the excavation of a protective ditch at the base of the talus (Fig. 21); and the erection of a wire-mesh fence as a further protective measure. Analysis of the 1974 rockfall with ROCKFALL, using similar model parameters to those used in the analysis of the Hedley and Sunnybrae incidents, simulates the path and run-out distance of the boulder (Fig. 22A). It also shows the effect of the ditch in stopping similar boulders (Fig. 22B) from moving into the rockfall shadow.

-1

I
SUNNYBRAE

I
0 200

Impact on house

400

600

metres

FIG. 19. Simulation of the Sunnybrae rockfall path using the computer program, with k , (normal restitution coefficient) = 0.70, kt (tangential restitution coefficient) = 0.80, and rolling friction coefficient = 0.52. Note position of actual stopping point of boulder.
ROCKFALL

Conclusions Rockfall is a common process in mountainous terrain where its long-term operation leads to the accumulation of talus slopes. Although deaths from rockfall in Canada are small in number, the indirect costs of land alienation because of rockfall hazard are very significant. In western Canada the costs of such land sterilization are in the order of tens of millions of dollars. As development pressures increase, techniques for assessing the vulnerability to big bo~ilderimpacts beyond the base of talus slopes have become a priority. A new terminology for use in rockfall hazard analysis is suggested. The area beyond the base of a talus slope that is reached by large boulders is termed the rockfall shadow, and the shadow angle is defined as the angle between the apex of the talus slope and the distal margin of the shadow. The angle between any point in the shadow and the apex of the talus slope is termed the equivalent shadow angle. Rockfall impact probability is a strong function of the distance from the source area. Within the rockfall shadow area return periods of the order of 1000 years relative to an individual house site are typical. Calculation of both landing and impact probabilities indicate that for the talus studied, a housing development located at 5 0 m from the

talus margin, covering an area 200 m wide along the talus base, would face a probability of being struck by a single rockfall boulder of 1 in 95 years. The prediction of rockfall behaviour in the shadow was geological, empirical, and analytic methcarried out ~ i s i n g ods. The evaluation of geological evidence can be helpful in assessing the probability of a defined shadow limit being violated and for establishing probabilities of impact within the shadow. An examination of 16 talus profiles from southwestern British Columbia yielded an empirical minimum shadow angle of 27.5", which is thought to correspond to the rollingfriction angle for the slope surfaces in question. This angle would be useful in a first approximation to shadow limits and is preferable to the use of the rockfall fahrboschung as proposed by several authors. The analytical lumped-mass model (ROCKFALL), which was developed, utilizes two restitution coefficients and a transition to rolling criterion. T h e model calculates the energy head of a boulder during its travel and was initially calibrated with reference to some full-scale Italian experiments. Although further calibration is necessary, results from the analysis of three rockfall incidents at the base of talus slopes in southwestern British Columbia were encouraging. In all cases the ROCKFALL simulation of boulder runout agreed almost exactly with that observed in the respective incidents. The documentation of the three rockfall incidents shows that, in each case, rockfall fragments impacted on homes

634

CAN. GEOTECH. J. VOL. 30, 1993

FIG. 20. View of Barnhartvale subdivision, Kamloops, British Columbia, in the vicinity of the 1974 rockfall, showing proximil of homes to talus slopes and source rock slopes. Arrows show protective ditch seen in Fig. 21.

FIG. 21. Protective ditch excavated following the 1974 Barnhartvale rockfall, photographed in 1981.

I1

. -

Topography

- Rockfall trajectory

i 1 -1

Impact on house

500 0 100 200 metres 300 400 0 100 200 metres 300 400

FIG. 22.

ROCKFALL

simulation of the Barnhartvale rockfall. (A) The 1974 rockfall. (B) The 1974 rockfall with protective ditch.

E V A N S A hI D HUNGR

635

at equivalent s h a d o w angles o f 30' o r more. T h i s would suggest that a review of existing development within rockfall shadow areas at the base of talus slopes may b e in order.

Acknowledgements
This paper is an outcome of joint research undertaken by the Geological Survey of Canada and Thurber Engineering Ltd. under the Government of Canada Unsolicited Proposal Program. T h e research was funded b y Supply and Services C a n a d a (Unsolicited Proposal Fund), Energy, M i n e s a n d Resources Canada (Geological Survey o f Canada) and Transport C a n a d a (Transportation D e v e l o p m e n t Centre). T h e authors acknowledge the assistance of t h e following people w h o contributed by providing helpful information, advice, and (or) encouragement related to this study in its various stages: M. Church, J.J Clague, D.M. Cruden, D. Fieber, D.R. Lister, J.C. Leighton, W.H. Mathews, G.C. Morgan, N.R. Morgenstern, G.E. Miller, H. Nasmith, L. Peckover, J.M. Ryder, N.A. Skermer, and P.J. Woods. T h e authors also wish t o thank Helen Moore, a survivor of the Hedley rockfall accident, for assistance in assembling data on the Hedley incident; and the Similkameen Indian Administration and C h i e f R o s s A l b e r t of t h e C o o k ' s F e r r y I n d i a n B a n d f o r allowing access to Indian lands. AstC, J.P., Cambou, B., and Falcetta, J.L. 1984. Comportement des masses rocheuses instables. De la prCvision 5 la prCvention. Iiz Proceedings of the 4th International Symposium on Landslides, Toronto, Ont. Vol. 1. pp. 441-446. AstC, J.P., Cambou, B., Clozel, P., and Falcetta, J.L. 1987. Quantitative analysis of cliff stability. IIZProceedings, International Symposium o n Engineering Geological Environment in Mountainous Areas, Beijing. Vol. 1. pp. 555-561. Azimi, C., and Desvarreux, P. 1977. Calcul de chutes de blocs et vkrification sur modkle rCduit. Association pour le developpement des recherches sur les glissements de terrain, Grenoble. Azimi, C., Desvarreux, P., Giraud, A,, and Martin-Cocher, J. 1982. MCthode de calcul de la dynamique des chutes de blocs. Application 5 1'Ctude du versant de la montagne de La Pale (Vercors). Bulletin de liaison des laboratoires des ponts et chaussCes, 122: 93-102. Azzoni, A., Rossi, P.P., Drigo, E., Giani, G.P., and Zaninetti, A. 1992. In situ observation of rockfall analysis parameters. In Landslides. Proceedings of the 6th International Symposium on Landslides, Christchurch, New Zealand, Edited b y D.H. Bell. A.A. Balkema, Rotterdam. Vol. 1. pp. 307-314. Ballivy, G., Hardy, J., Lebuis, J., and Bergeron, R. 1984. Techniques d'abattage d'une paroi rocheuse instable. In Proceedings of the 4th International Symposium on Landslides, Toronto, Ont. Vol. 1. pp. 455-459. Brundsen, D., Jones, D.K.C., and Goudie, A.S. 1984. Particle size distribution on the debris slopes of the Hunza Valley. Irz The International Karakoram Project. Edited b y K.J. Miller. Cambridge University Press, Cambridge, U.K. Vol. 2. pp. 536-580. Boyd, R.D. 1991. Rockfall containment measures at Springdale, Newfoundland. In Slope stability engineering; developments and applications. Edited b y R.J. Chandler. Thomas Telford Ltd., London. pp. 409-4 13. Bozzolo, D., and Pamini, R. 1986. Simulation of rock falls down a valley side. Acta Mechanics, 63: 113-130. Bozzolo, D., Pamini, R., and Hutter, K. 1988. Rockfall analysisa mathematical model and its test with field data. Iiz Landslides. Proceedings of the 5th International Symposium on Landslides, Lausanne. Edited b y C . Bonnard. A.A. Balkema, Rotterdam. Vol. 1. pp. 555-560.

Broili, L. 1974. Ein Felssturz im Grossversuch. Rock Mechanics, L Supplement 3, pp. 69-78. Broili, L. 1977. Relations between scree slope morphometry and dynamics of accumulation processes. Istituto Sperimentale Modelli e Strutture, Bergamo. Publication No. 90. pp. 11-23. Church, M., Stock, R.F., and Ryder, J.M. 1979. Contemporary sedimentary environments on Baffin Island, N.W.T., Canada: Debris slope accumulations. Arctic and Alpine Research, 11: 371-402. Clague, J.J. 1980. Late Quaternary geology and geochronology of British Columbia. Part 1: Radiocarbon dates. Geological Survey of Canada, Ottawa. Paper 80-13. Descoeudres, F., and Zimmerman, T.H. 1987. Three-dimensional dynamic calculation of rockfalls. In Proceedings of the 6th International C o n g r e s s o n R o c k Mechanics, MontrCal, Que. A.A. Balkema, Rotterdam. Vol. 1. pp. 337-342. Eisbacher, G.H., and Clague, J.J. 1984. Destructive mass movements in high mountains; hazard and management. Geological Survey of Canada, Ottawa. Paper 84- 16. Falcetta, J.L. 1985. Un nouveau modble de calcul de trajectoires de blocs rocheux. Revue Franqaise de GCotechnique, 30: 11-17. Fenti, V., Silvano, S., and Spagna, V. 1977. Rockfalls in the Isarco valley (South Tyrol) and proposed methods for a provisional map. Instituto Sperimentale Modelli e Strutture, Bergano. Publication No. 90. pp 55-65. Fornaro, M., Peila, D., and Nebbia, M. 1990. Block falls on rock slopes-application of a numerical simulation to some real cases. In Proceedings of the 6th International Association of Engineering Geology Congress, Amsterdam. A.A. Balkema, Rotterdam. Vol. 3. pp. 2173-2180. Gardner, J.S. 1970. Rockfall: a geomorphic process in high mountain terrain. Albertan Geographer, 6: 15-20. Gardner, J.S. 1980. Frequency, magnitude, and spatial distribution of mountain rockfalls and rockslides in the Highwood Pass area, Alberta, Canada. In Thresholds in geomorphology. Edited b y D.R. Coates and J.D. Vitek. Allen and Unwin, London. pp. 267-295. Govi, M. 1977. Photo-interpretation and mapping of the landslides triggered by the Friuli Earthquake (1976). Bulletin of the International Association of Engineering Geology, 15: 67-72. Govi, M., and Sorzana, P.F. 1977. Effeti geologici del terremoto: frane. Rivista Italiana di Paleontologia e Stratigrafia, 83: 329-368. Heim, A. 1932. Bergsturz und Menschenleben. Beiblatt zur Vierteljahrsschrift der Naturforschenden Gesellschaft in Ziirich, 77: 1-217. (English translation by N.A. Skermer, Bitech Publishers, Vancouver.) Hestnes, E. 1980. Skredfarevurdering. Norwegian Geotechnical Institute, Oslo. Publication 132. pp. 61-81. Hungr, O., and Evans, S.G. 1988. Engineering evaluation of fragmental rockfall hazards. 111 Landslides. Proceedings of the 5th International Symposium on Landslides, Lausanne. Edited b y C. Bonnard. A.A. Balkema, Rotterdam. Vol. 1. pp. 685-690. Hungr, O., and Evans, S.G. 1989. Engineering aspects of rockfall hazards in Canada. Geological Survey of Canada, Ottawa. Open File 2061. Hutchinson, J.N. 1988. Morphological and geotechnical parameters of landslides in relation to geology and hydrogeology, state-of-the-art report. In Landslides. Proceedings of the 5th International Symposium on Landslides, Lausanne. Edited b y C. Bonnard. A.A. Balkema, Rotterdam. Vol. 1. pp. 3-35. Kirkby, M.J., and Statham, I. 1975. Surface stone movement and scree formation. Journal of Geology, 83: 349-362. Kobayashi, Y., and Kagawa, T. 1987. The prediction of hazards from debris avalanches and rockfalls with the aid of computer simulations. Iiz Proceedings of the International Symposium on Engineering Geological Environment in Mountainous Areas, Beijing. Vol. 1. pp. 563-572. Kobayashi, Y., Harp, E.L., and Kagawa, T. 1990. Simulation of

636

CAN. GEOTECH. J . VOL. 30, 1993

rockfalls triggered by earthquakes. Rock Mechanics and Rock Engineering, 23: 1-20. Kolderup, N.-H. 1955. Raset I Modalen 14. August 1953. Norsk Geologisk Tidsskrift, 34: 21 1-217. Kotarba, A., and Stromquist, L. 1984. Transport, sorting and deposition processes of alpine debris slope deposits in the Polish Tatra Mountains. Geografiska Annaler, Series A, 66: 285-294. Lied, K. 1977. Rockfall problems in Norway. Istituto Sperimentale Modelli e Strutture, Bergamo. Publication No. 90. pp 51-53. Luckman, B.H. 1988. Debris accumulation patterns on talus slopes in Surprise Valley, Alberta. GCographie physique et Quaternaire, 42: 247-278. McConnell, R.G., and Brock, R.W. 1904. The great landslide at Frank, Alberta. Geological Survey of Canada, Ottawa. Annual Report 1902-1903, Part 8, Appendix. Miller, G.E. 1983. Rockfall investigation, Bastion Mountain, Sunnybrae, B.C. British Columbia Ministry of Environment, Victoria, B.C. Onofri, R., and Candian, C. 1979. Indagine sui limiti di massima invasione dei blocchi rocciosi franati durante il sisma del Fruili del 1976. Considerazioni sulle opere di difesa. Regione Autonoma Friuli-Venezia Giulia, Cluet, Trieste. Paronuzzi, P. 1988. Masse rocciose in mot0 su pendii naturali: aspetti cinematici e dinamica. Bolletino della SocietB Geologica Italians, 107: 153-170. Paronuzzi, P. 1989. Probabilistic approach for design optimization of rockfall protective barriers. Quarterly Journal of Engineering Geology, 22: 175-183. Pfeiffer, T.J., and Bowen, T.D. 1989. Computer simulation of rockfalls. Bulletin of the Association of Engineering Geologists, 26: 135-146. Piteau, D.R., and Clayton, R. 1977. Discussion of paper "Computerized design of rock slopes using interactive graphics for the input and output of geometrical data" by P.A. Cundall, M.D. Voegele, and C. Fairhurst. In Design Methods in Rock Mechanics, Minneapolis. Proceedings of the 16th Symposium on Rock Mechanics. Edited by C. Fairhurst and S.L. Crouch. American Society of Civil Engineers, New York. pp. 62-63. Rapp, A. 19600. Talus slopes and mountain walls a t Tempelfjorden, Spitsbergen. Norsk Polarinstitutt Skrifter, 119: 1-96.

Rapp, A. I960b. Recent development of mountain slopes in KHrkevagge and surroundings, northern Scandinavia. Geografiska Annaler, 42: 65-200. Ritchie, A.M. 1963. Evaluation of rockfall and its control. Highway Research Board Record, 17: 13-28. Rochet, L. 19870. Application des modkles numCriques de propagation B 1'Ctude des Cboulements rocheux. Bulletin de liaison des laboratoires des ponts et chaussCes, 150-151: 84-95. Rochet, L. 19876. DCveloppement des modkles numCriques dans l'analyse d e la propagation d'Cboulements rocheux. I n Proceedings of the 6th International Congress on Rock Mechanics, Montreal, Que. A.A. Balkema, Rotterdam. Vol. 1. pp. 479-484. Spang, R.M. 1987. Protection against rockfall-stepchild in the design of rock slopes. In Proceedings of the 6th International Congress on Rock Mechanics, MontrCal, Que. A.A. Balkema, Rotterdam. Vol. 1. pp. 551-557. Spang, R.M., and Rautenstrauch, R.W. 1988. Empirical and mathematical approaches to rockfall protection and their practical applications. I n Landslides. Proceedings of the 5th Symposium on Landslides, Lausanne. Edited by C. Bonnard. A.A. Balkema, Rotterdam. Vol. 2. pp. 1237-1243. Statham, I., and Francis, S.C. 1986. Influence of scree accumulation and weathering on the development of steep mountain slopes. In Hillslope processes. Edited by A.D. Abrahams. Allen and Unwin, London. pp. 245-267. Toppe, R. 1987. Terrain models-a tool for natural hazard mapping. In Avalanche formation, movement and effects. Edited by B. S a l m and H . Gubler. International Association of Hydrological Sciences, Wallingford, U.K. Publication 162. pp. 629-638. Varnes, D.J. 1978. Slope movement types and processes. I n Landslides, analysis and control. Edited by R.L. Schuster and R.J. Krizek. Transportation Research Board, National Research Council, Washington, D.C. Special Report 176. pp. 11-33. Worobey, G.A. 1972. An investigation of talus slope development in the Similkameen Valley, near Keremeos, British Columbia. M.A. thesis, University of British Columbia, Vancouver. Wu, S.-S. 1985. Rockfall evaluation by computer simulation. Transportation Research Record, 1031: 1-5.