Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

Engineering Geology 73 (2004) 247 – 265

www.elsevier.com/locate/enggeo

A warning system for rainfall-induced shallow failures


Pietro Aleotti *
Aquater S.p.A., Via Tolstoj, 86 20098, San Giuliano Milanese, Italy

Abstract

It is widely recognised that soil slips and debris flows are triggered by short intense storms. Owing its geologic,
geomorphologic and climatic settings, the Piedmont Region (NW Italy) is highly prone to the occurrence of this kind of
landslides. In the last two centuries, in fact, a total of 105 severe meteoric events which triggered shallow failures occurred and,
of these, 18 events took place from 1990 to 2002. A fair number of rainfall thresholds have been proposed in the literature,
defined both on empirical or on physical bases. Empirical thresholds are defined collecting rainfall data for landslide meteoric
events and for events without landslides, while physical thresholds are based on numerical models that consider the relation
between rainfall, pore pressure and slope stability. The main objective of this paper is the identification of the empirical
triggering thresholds for the Piedmont Region. Four meteoric events were selected and analysed (November 4 – 5, 1994; July
7 – 8, 1996; April 27 – 30, 2000; October 13 – 16, 2000) because they supply a wide range of variation for both rainfall
parameters (duration, intensity, cumulative rainfalls) and the number of induced landslides. In the intensity – duration plot, the
critical limit is described by the equation: I = 19D 0.50 (where I = rainfall intensity expressed in mm/h and D = rainfall duration
expressed in hours). Such a limit is traced to envelop 90% of the points on the graph. In the NI – D diagram the triggering
thresholds are given by the equations NI = 0.76D 0.33 and NI = 4.62D 0.79 (where NI = normalised intensity with respect to the
annual precipitation, MAP, expressed in %, [(mm/h)/PMA]  100). In the article the different meaning of these thresholds is
discussed. Finally, the diagram NI – NCR is proposed; the triggering threshold is given by the expression:
NI =  0.09ln[NCR] + 0.54 (where NCR is the normalised cumulative critical rainfall, [mm/PMA]  100). The application of
the triggering thresholds as a fundamental element in a warning system dedicated to the safeguarding of population in landslide-
prone areas is discussed. In detail an operating procedure which is presently being verified and tested in the studied area is
described.
D 2004 Elsevier B.V. All rights reserved.

Keywords: Shallow failures; Soil slips; Debris flows; Rainfall; Triggering thresholds; Warning system; Piedmont Region (Italy)

1. Introduction posing elements, and above all, as triggering factors


(Wieczorek, 1996).
The factors affecting slope stability are various, and The influence of rainfall on landslides differs sub-
virtually most are closely interconnected. Among stantially depending upon landslide dimensions, kine-
these, meteoric events (or rather, the effects induced matics, material involved, etc. Shallow failures are
by these) are of primary importance, both as predis- usually triggered by short intense storms (Campbell,
1975; Lumb, 1975; Brand et al., 1984; Cancelli and
* Tel.: +39-2-52052600; fax: +39-2-52052586. Nova, 1985; Cannon and Ellen, 1985; Wieczorek,
E-mail address: pietro.aleotti@aquater.eni.it (P. Aleotti). 1987; Guzzetti et al., 1992; Polloni et al., 1992; Morgan

0013-7952/$ - see front matter D 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.enggeo.2004.01.007
248 P. Aleotti / Engineering Geology 73 (2004) 247–265

et al., 1997; Crosta, 1998; Corominas and Moya, 1999; This paper discusses soil slips and soil slip-debris
Flentje et al., 2000; Paronuzzi et al., 2002) while most flows in the Piedmont Region, North West Italy (Fig.
deep-seated landslides are affected by long-term vari- 1): due to its physical setting (geology, lithology,
ation of annual rainfall which has to last several years climatic conditions) this region is highly prone to this
(Bonnard and Noverraz, 2001). For landslides such as type of slope instability (Anselmo, 1980; Govi et al.,
rockfalls, no precise correlation with rainfall can be 1985; Aleotti et al., 1996; Bandis et al., 1996; Polloni
surmised, as they appear more sensitive towards other et al., 1996; Sutera Sardo et al., 1996; Aleotti et al.,
factors such as chemical –mechanical weathering of the 1998; Regione Piemonte, 1998; Aleotti and Polloni,
rockmass and temperature fluctuation across the freez- 2000; Aleotti et al., 2002). The triggering thresholds
ing point (Sandersen et al., 1996): only late spring and determined for this type of failures are then applied
summer rockfalls can be related with rainfall. For within a warning system.
example, a large fall occurred, after some days of
intense precipitation, in the Brembana Valley (Italian
Central Alps) at the beginning of May 2002, killing 2. Landslide-triggering rainfall thresholds
three persons.
Based on the above considerations, it derives that Generally speaking, the term ‘‘threshold’’ defines
there are no critical rainfall conditions for every type the minimum or maximum (critical) level of some
of landslide. quantity needed for a process to take place (Reich-

Fig. 1. Location of the studied area (Piedmont Region, NW Italy). Patterns indicate the distribution of areas affected by events of shallow
landsliding (November 1994, July 1996, April 2000, June 2000, October 2000). Regional layout.
P. Aleotti / Engineering Geology 73 (2004) 247–265 249

enbach et al., 1998). Since the objective of this study is statistical) (Campbell, 1975; Caine, 1980; Crozier
to define a rainfall threshold with practical applications and Glade, 1999);
in warning systems, thresholds of interest are the (ii) Physical thresholds, based on numeric models that
minimum ones, i.e. thresholds that define the lowest take into account the relationship between rainfall,
level above which one or more than one landslide can pore pressure and slope stability by coupling
be triggered. hydrologic and stability models (Montgomery and
An analytical dissertation of the concept of rainfall Dietrich, 1994; Wilson and Wieczorek, 1995;
threshold was recently presented by Iiritano et al. Crosta, 1998; Terlien, 1998).
(1998) who introduced the so-called mobilisation
function, Y(t), that indirectly describes all the factors Physical thresholds are not widely developed and,
that contribute to trigger a landslide and that is generally, they require detailed knowledge of the
dependent, at every time t, on the amount of water boundary conditions, which are seldom available
infiltrated in the soil before the time t (Eq. (1)): outside specially equipped test fields (rain gauges,
piezometers, tensiometers). Recent attempts at a re-
Y ðtÞ ¼ f ½IðuÞ with  l < uVt ð1Þ gional scale have been proposed by using distributed
models (Borga et al., 1998; Aleotti et al., 2003; Crosta
where I(u) is infiltration intensity at time u.
et al., 2003).
By defining P[Lt] as the probability of occurrence
On the contrary, there are a fair number of empirical
of a landslide at time t, and assuming Y1 as the
thresholds and different graphs have been used to
mobilisation coefficient over which the probability
represent them, depending upon combinations of the
of landslide occurrence is different from zero, and
most commonly used rainfall parameters: antecedent
Y2 as the mobilisation coefficient over which the
rainfall, duration, intensity, cumulative rainfall. In the
probability of landslide occurrence is certain, the
first suggested elaboration, the triggering thresholds
following alternative hypothesis can be derived:
are simply defined either by the critical cumulative
P½Lt  ¼ 0 if Y ðtÞ < Y1 ð2aÞ rainfall (Campbell, 1975) or by the rainfall intensity
(Brand et al., 1984). The most commonly used thresh-
P½Lt  ¼ g½Y ðtÞ if Y1 VY ðtÞVY2 ð2bÞ olds are those defining the intensity– duration (Caine,
1980), although some conceptual differences among
P½Lt  ¼ 1 if Y ðtÞ > Y2 ð2cÞ the various authors exist. Firstly, the dimensions of the
study area vary from a few square kilometres (Wiec-
where g[ Y(t)] is a generic function defined in the zorek, 1987), to large regions (i.e. Austrian Alps,
interval [ Y1;Y2] having co-domain [0,1]. Eq. (2a) Moser and Hohensinn, 1983). Then, the relationship
indicates impossible mobilisation due to rainfall (i.e. suggested by Cannon and Ellen (1985) attempts to
unconditionally stable), vice versa Eq. (2c) indicates predict the occurrence of ‘‘abundant landslides’’, while
certain mobilisation (i.e. unconditionally unstable). that proposed by Wieczorek (1987) only ‘‘one or more
Assuming Y1 = Y2 = YTH, a ‘‘threshold’’ scheme is than one landslide’’. The ‘‘intensity – duration’’ ap-
introduced, free from any probabilistic connection, in proach can be further refined by normalising the
which the event Lt can be considered alternatively as intensity value with the mean annual rainfall (MAP),
impossible (Eq. (3a)) or certain (Eq. (3b)): thus emphasising the regionalization of the thresholds,
since the calculation takes into account the climatic
P½Lt  ¼ 0 if Y ðtÞVYTH ð3aÞ regimes of the study area (Cannon and Ellen, 1985;
P½Lt  ¼ 1 if Y ðtÞ > YTH ð3bÞ Jibson, 1989; Wieczorek et al., 2000).
Another frequently used graph correlates the total
Two types of landslide-triggering rainfall thresh- amount of rainfall until landslide occurrence (critical
olds can be established: cumulative rainfall) with the maximum recorded inten-
sity. A similar representation was proposed by Govi et
(i) Empirical thresholds, based on historic analysis of al. (1985) in which, however, the critical rainfall value
relationship rainfall/landslide occurrence (i.e. is compared to mean annual precipitation.
250 P. Aleotti / Engineering Geology 73 (2004) 247–265

Antecedent rainfall conditions can be used to of the landslides occurrence), and also by the need to
identify thresholds associated to critical daily rainfall. examine events that can supply the widest possible
In the so-called Antecedent Daily Rainfall Model range of variation for significant parameters such as
(Glade et al., 2000), the antecedent rainfall (10 days) duration, intensity, critical rainfalls (mean and maxi-
is connected with critical one after being reduced by mum) and number of induced landslides. Fig. 2
a decay factor that defines the rate of soil moisture summarises for each events the cumulative rainfall
decrease within a specific period. The Antecedent trend. The pluviometric characteristics of critical rain-
Soil Water Status Model (Glade, 2000b) calculates falls measured during the events have a rather large
the interaction between antecedent rainfall, soil mois- spectrum, both in terms of duration (between 7 and
ture and potential evapotranspiration, and gives as a 112 h) and (mean) critical intensity (between a min-
result the probability of landslide occurrence for a imum of 2 mm/h and a maximum of over 26 mm/h).
given combination of daily rainfall magnitude and The number of triggered landslides differs substan-
soil moisture, expressed as the soil water status tially from one situation to another as well as the
index. density.
The November 4– 5, 1994 event affected a very
large territory (7500 km2) and was probably the most
3. Rainfall events in the piedmont region catastrophic of the 20th Century in Piedmont, both in
terms of damage (80 casualties and approximately
The Piedmont Region is frequently affected by 1  109 US$ of damage) and of number of landslides
severe meteoric events. From 1800 to 1990, a total (more than 10,000, many as new failures). The event
of 87 episodes occurred, which triggered landslides. aroused great interest in the Italian scientific commu-
Events are mainly concentrated in autumn (53%) that nity and resulted in a large number of publications
is the most rainy season in this region (Luino et al., (Aleotti et al., 1996; Bandis et al., 1996; Clarizia et al.,
1994). 1996; Forlati et al., 1995; Polloni et al., 1996; Sutera
In addition, numerous events took place recently in Sardo et al., 1996). The present study focuses on the
this region, the most significant being: 3rd – 5th Oc- Langhe Cuneesi area where soil slips were most
tober 1992, 22nd –24th September 1993, 18th May abundant (approximately 3000 with density higher
1994, 26th June 1994, 23rd September 1994, 4th – 5th than 100 landslides/km2 in the upper portion of
November 1994, 7th –8th July 1996, 8th –9th Octo- secondary basins).
ber 1996, 28th – 29th June 1997, 28th August 1997, The centre of the July 7 – 8, 1996 flood was a small
4th – 5th September 1998, 23rd October 1999, 27th – area (70 km2) near Mt. Mottarone. The storm trig-
30th April 2000, 14th –16th June 2000, 30th Septem- gered around 30 soil slips and debris flows in four
ber 2000, 13th –16th October 2000, 2nd – 5th May watersheds: debris torrents spread out on the alluvial
2002, 4th –6th June 2002. fans, causing one victim and severe damage (Chiarle
In this study the following events are analysed and Luino, 1996).
(Fig. 1): (i) November 1994 in Langhe Cuneesi area; The April 27 – 30, 2000 meteorologic event in-
(ii) July 1996 in Cusio-Verbano area (Mt. Mottarone); volved the mountain and foothill areas of the western
(iii) April 2000 in western Piedmont; (iv) June 2000 hydrographic basins, from the Po to the Dora rivers. It
in south-western Piedmont; (v) October 2000 in originated 100 soil slips, some of which evolved to
western and northern Piedmont. debris flows. In some areas the soil slips started as
The choice was dictated by the need to examine translational slides. Man activities represented an
events for which a greater amount of data was indirect contributory cause of failure in almost 30%
available (including landslide type, location and time of the recorded cases: they are due, particularly, to

Fig. 2. Cumulative rainfalls versus time (15 days antecedent and critical rainfalls) recorded at various rain gauges in the areas affected by the
selected meteoric events: (a) November 1994; (b) July 1996; (c) April 2000; (d) June 2000; (e) October 2000 (Orco Valley, Sesia Valley, Stura
Valley, Ossola Valley). Small square in the left-hand side shows an example of ‘‘time lag’’ in the beginning of critical rainfall in different rain
gauges during the November 1994 meteoric event.
P. Aleotti / Engineering Geology 73 (2004) 247–265 251
252 P. Aleotti / Engineering Geology 73 (2004) 247–265

road cuts/embankment on hill slopes, inadequate


drainage of surface water and overloading.
The June 14 –16, 2000 event took place in western
and south-western Piedmont, and affected mainly the
area between the Susa Valley (Dora River), to the
north, and Pesio Valley, to the south. More than 100
landslides (102) were inventoried. The vast majority
of failures (95) were soil slips and occurred primarily
in the Maira and Stura di Demonte valleys (almost
40% of total landslides), causing interruptions and
damages to roads (82%) and buildings (10%)
(Regione Piemonte, 2000).
The October 13 –16, 2000 meteoric event struck
Fig. 3. Definition of rainfall parameters.
approximately 25% of the entire Piedmont Region.
Almost 500 landslides were triggered, especially in
the northern (Ossola Valley, Sesia Valley, Lanzo and Duration of the critical rainfall event, always ex-
Orco valleys) and western parts (Pellice Valley) of the pressed in hours in the processing, is the time that
region. Over 62% of the landslides consisted in soil elapses from the beginning of critical precipitation to
slips, and 19% were rotational slides. Falls (5%) and activation of the landslides. In this paper rainfall data
other types of landslides were less frequent (Aleotti et are normalised with respect to the mean annual
al., 2002). Most of the material involved by the precipitation (MAP). The ratio between the critical
landslides consisted of incoherent deposits, and to a precipitation of the event and the mean annual pre-
lesser extent, of debris and loose soil. Almost every- cipitation of the site is defined as Normalised Critical
where, the most consistent damage caused by slope Rainfall (NCR). In literature this ratio is also called
failures were along primary and secondary roads Normalised Storm Rainfall (NSR) (Guidicini and
(70%) and, secondly, in buildings (16%). Iwasa, 1977).
The Piedmont Region is monitored by a great
number of rain gauges. Selection of the rain gauge
4. Rainfall thresholds that will be associated to a slide is extremely impor-
tant, as the pelt-point responsible for superficial slides
4.1. Input data can be both small and rapid. This may involve a
significant difference in data registered by two close
In a recent publication, Crosta and Frattini (2001) gauges for the same rainfall episode. The difference
emphasised the importance of having accurate data- does not concern normalised critical precipitation
bases containing all the information needed to study values only (see Fig. 2), but also how rainfalls
rainfall events. Equally important is the type of develop, i.e. the beginning of critical rainfall. For
information used in a study, which must be carefully example, if the beginning of critical precipitation is
defined in order to enable comparison between data not carefully identified, it could lead to misleading
produced by researchers operating in very different quantification of critical rainfall values and mean
geographic contexts. intensities and hence, incorrect calculation of critical
In this study we used hourly rainfall data for thresholds. It has been observed that during a meteoric
critical precipitation, and daily rainfall measurements event within a fairly small area, the inset of critical
for the previous days. Conventionally, critical precip- rainfall can vary substantially from one gauge to
itation indicates the amount of rainfall from the time another (even 12 and 15 h in the June 2000 and
(‘‘zero point’’) in which a sharp increase in rainfall October 2000 floods, respectively).
intensity is observed and the triggering of the (first) Fig. 4 shows the importance of the location of rain
landslide (Fig. 3). This increase results in a significant gauges, which should be as close as possible to the
break in the slope of the rainfall cumulative curve. landslide. This graph was obtained by measuring the
P. Aleotti / Engineering Geology 73 (2004) 247–265 253

Fig. 4. Correlation between the mutual distance of rain gauges and maximum difference in normalised hourly rainfall.

mutual distance between the gauges in a given area Passo Moro (maximum hourly intensity 7.6 mm/h). In
and by calculating, for each pair of gauges, the this case the difference can be attributed both to the
maximum difference measured in hourly precipitation elevation and to the position of the rain gauges: the
(normalised) during an event of intense rainfall. We one in Macugnaga Passo Moro is located at 3000 m
can reasonably conclude that, even with a significant above the sea level and near a crest of the watershed
scatter (r2 = 0.2), a direct relation exists between the and the one in Macugnaga Pecetto is situated at 1500
maximum difference in rainfall values and the dis- m a.s.l. on a large slope that might act as a barrier
tance between rain gauges. Such a relation is greater against cloud fronts.
for distances lower than 10– 15 km and decreases for
larger distances. 4.2. Relationship between antecedent and critical
Additional research is needed to determine the rainfall
importance of selecting a reference gauge for a given
landslide. The significance of rainfall data referred to It has been recognised in the literature that anteced-
a certain surficial slides does not depend entirely on ent rainfall can be a predisposing factor in the activation
distance, but also on other factors such as elevation of soil slips (Wieczorek, 1987). The influence of
and aspect of the measuring gauge, and generally, on antecedent rainfall is difficult to quantify as it depends
its position (also in relation to the prevailing wind on several factors, including the heterogeneity of soils
direction in the area). (strength and permeability properties) and the regional
A significant example is the difference between the climate. In tropical areas, for example, antecedent
values recorded by the rain gauges in Macugnaga rainfall is not an important factor (Brand, 1992) as well
Passo Moro and Macugnaga Pecetto during the Oc- as in slopes covered with coarse colluvium having large
tober 2000 flooding event in Ossola Valley (Fig. 5). interparticle voids, debris flows can occur without
Although the distance between the two gauges is only significant antecedent rainfall (Corominas and Moya,
1.5 km, from 12 to 16 October 2000 the rain gauge in 1999). Similar response can be observed even in
Macugnaga Pecetto measured 576 mm of rainfall pervious soils due to the presence of preferential
(maximum hourly intensity 16.2 mm/h) as opposed groundwater passageways (macropores, animal bur-
to the 158 mm measured by the gauge in Macugnaga rows, root channels) (Corominas, 2000). On the con-
254 P. Aleotti / Engineering Geology 73 (2004) 247–265

Fig. 5. Difference in rainfall values measured by rain gauges in Macugnaga Pecetto and Macugnaga Passo Moro during the October 2000
meteoric event: bars correspond to the hourly rainfalls (left Y-axis, mm/h) recorded in Macugnaga Pecetto rain gauge (white bars) and
Macugnaga Passo Moro rain gauge (black bars), lines indicate the relevant cumulative rainfalls (right Y-axis, mm).

trary, in low-permeability soils antecedent rainfall can 4.3. Empirical thresholds


be an important factor because it reduces soil suction
and increases the pore-water pressures in soils. How- A preliminary assessment consisted in calculating
ever, the time interval established as significant by the the rain path measured at various rain gauges during
various authors differs considerably: 5 days (Wiec- the events, and plotting it on intensity – duration
zorek et al., 2000), 10 days (Crozier, 1999; Glade et graph. The rain paths were calculated by establishing
al., 2000), 15 days (Govi et al., 1985). the initial hour of a given critical rainfall in a selected
The time intervals taken into consideration for the gauge, and determining the mean intensity at fixed
events examined in this paper were 7, 10 and 15 days. intervals (1, 3, 6, 12 and 24 h from start) until the first
However, results show that not significant correlation landslide was triggered. Thus, the final point repre-
exist between antecedent and critical rainfall: the sents the mean intensity calculated over the entire
scattering of the population sample, in fact, is very period of critical rainfall.
high and the relevant correlation coefficients are The rain path indicates that in the majority of cases
r2 = 0.17, r2 = 0.16 and r2 = 0.32 for the 7-, 10- and (48%) failure occurs when the 50-year return period
15-day intervals, respectively. Curves in Fig. 6 envel- of rainfall is exceeded (Fig. 7). This limit is more
op 90% of the plotted rainfall events for 7- and 10-day significant in the case of exceptional rainfall events, as
intervals. Equations are expressed as follows: in the floods of November 1994 (64%) and October
2000 (53%). Instead, for less intense rainfall episodes
NCR ¼ 12:45e0:15NAR ð4Þ (April 2000), the RT10 limit becomes the triggering
threshold (71%).
Based on the above considerations we can con-
NCR ¼ 11:50e0:08NAR ð5Þ clude that the curves calculated in the rain gauges for
rainfalls with different return periods (RT10, RT50)
where NCR = normalised critical rainfall (%) and may represent a first type of triggering thresholds,
NAR = normalised antecedent rainfall (%). which despite the approximation could represent ef-
P. Aleotti / Engineering Geology 73 (2004) 247–265 255

Fig. 6. Relation between normalised antecedent rainfalls (NAR) and normalised critical rainfalls (NCR) for 7- and 10-day intervals. Lines
envelop 90% of the plotted data for 7- and 10-day intervals.

fective limits for critical rainfall events of varying capable of triggering numerous landslides (as the
magnitude. The RT10 threshold could be representa- November 1994 event). Notwithstanding the use of
tive of rainfall events that cause a modest number of these thresholds in a warning system is not advisable.
failures (low magnitude, such as the April 2000 In the intensity –duration graph, we identified a
event), while the RT50 threshold could indicate events critical limit for fairly wide range of durations (Fig. 8).

Fig. 7. Position of failures with respect to the statistical curves RT10 and RT50.
256 P. Aleotti / Engineering Geology 73 (2004) 247–265

Fig. 8. Triggering threshold in graph I – D for shallow landslides in the Piedmont Region. Curve [90] encompasses 90% of the available data and
represents the triggering threshold; the other limits [75, 50, 50, 25, 10] envelop different percentage of points in the graph.

The equation of this threshold (curve [90]) assumes on is also very close to the classical Caine threshold
the following exponential law (Eq. (6)): (1980): this similarity is even more surprising consid-
ering that the threshold suggested by Caine has been
I ¼ 19D0:50 ð6Þ calculated as a valid limit for the entire World.
By normalising the intensity value, it is possible to
where I=(mean) rainfall intensity calculated in mm/ represent the three rainfall parameters simultaneously
h and D = duration of critical rainfall event expressed (intensity, duration and mean annual rainfall) (Can-
in hours. non, 1988) (Fig. 10). The points indicate the maxi-
This curve is calculated to include 90% of the mum intensities registered for the duration considered
points on the graph; the 90% limit was established (1, 3, 6, 12 and 24 h), besides the normalised intensity
arbitrarily and is dictated by the need to both limit the relative to the moment in which the landslide was
maximum number of cases possible, and also, to triggered.
eliminate sporadic and therefore unrepresentative Fig. 10 shows two different thresholds in graph
cases. The other curves of the graph were calculated NI – D (Eqs. (7) and (8)), which refer to two different
for inclusions 75%, 50%, 25% and 10%, respectively, methods of data interpolation.
and take on the following intercept values: 23.5, 34.0,
52.3 and 64.9. In graph I– D the critical threshold was NI ¼ 0:76D0:33 ðcurve AÞ ð7Þ
represented as a line rather than a curve since the
intense and short events (shorter than 10 h) are not NI ¼ 4:62D0:79 ðcurve BÞ ð8Þ
adequately represented in the set of available data.
When the new threshold is compared with those where NI = normalised critical intensity (%) and
available in literature (Fig. 9), it appears very similar D = duration of critical event (h).
to the one proposed for Italian Central Alps (Ceriani et Curve A encompasses all the available data in
al., 1994). The difference is a slightly lower exponent figure, considering critical rainfall in its whole (Can-
(  0.50 against  0.55). It is interesting to note that it non and Ellen, 1985). Curve B considers only nor-
P. Aleotti / Engineering Geology 73 (2004) 247–265 257

Fig. 9. Comparison of the triggering threshold in graph I – D with those reported in the literature. The thresholds proposed by Caine (1980) and
Crosta and Frattini (2001) are prepared by using all the data available at the date for the world; the thresholds by Cancelli and Nova (1985),
Ceriani et al. (1994) and Clarizia et al. (1996) are referred to wide areas with different soil, morphologic and rainfall characteristics. Threshold
proposed by Bolley and Oliaro (1999) is calculated for the upper Susa Valley (Dora river, see Fig. 1) within the studied area of this paper.

Fig. 10. Triggering thresholds in graph NI – D for the Piedmont Region. Curve A encompasses all the available data, while curve B considers
only normalised intensity relative to the moment in which the landslide was triggered.
258 P. Aleotti / Engineering Geology 73 (2004) 247–265

Fig. 11. Definition of critical triggering thresholds in NI – NCR graph. Curve [100] envelops all the available data; curve [90] encompasses 90%
of the points in the graph.

malised intensities relative to the moment in which the rainfall –landslide thresholds. One of the first warning
landslide was triggered, thus enveloping the less systems was developed by the USGS in the San
dispersed part of the cluster of points. A comparison Francisco Bay area (Keefer et al., 1987; Wilson and
with the thresholds available in literature for different Wieczorek, 1995). It was based on the quantitative
territorial contexts shows a certain similarity with the precipitation rainfall forecast (QPRF) provided by the
curves proposed by Ceriani et al. (1994) and Wilson et National Weather Service (each forecast gave pre-
al. (1992). dicted rainfall for the upcoming 6 h), on a network
The curve identified as critical threshold in the of more than 40 real-time continuous rainfall gauges,
diagram NCR – NI (RBMCJ, 1985) is given by Eq. (9): and on the rainfall threshold for initiation of landslid-
ing (Cannon and Ellen, 1985). When actual real-time
NI ¼ 0:09lnðNCRÞ þ 0:54 ð9Þ readings in conjunction with predicted values
approached the threshold, a warning message was
where NI = normalised intensity (%) and NCR = nor- issued. Similar systems were also developed and
malised critical rainfall (%) (Fig. 11). Curve (9) corre- described in Hong Kong (Brand et al., 1984), Italy
sponds to the lower limit of 90% of the points on the (Sirangelo and Braca, 2001), Japan (Onodera et al.,
graph. 1974), New Zealand (Crozier, 1999), South Africa
(Gardland and Olivier, 1993) and Virginia (Wieczorek
and Guzzetti, 1999). In Hong Kong, the Geotechnical
5. Applying thresholds to warning systems Engineering Office (GEO) has recently implemented a
new entirely automated computer system (Landslip
An interesting use of empirically based thresholds Warning System) that is the world first in landslide
is in warning systems. Indeed, thresholds are an forecasting and is now in everyday use (Premchitt,
important element of these systems, which must be 1997). The Landslip Warning System is based on
composed of several basic components related to short-term rainfall forecast and it is composed of 86
rainfall forecasts, real-time rainfall monitoring and rain gauges. In addition to data from rain gauges,
P. Aleotti / Engineering Geology 73 (2004) 247–265 259

radar and satellite images are used to monitor the (curve A in Fig. 12), or as curves that, once the critical
movement and development of rain-bearing clouds. time Dtc has been established (i.e. the minimum time
The Landslip Warning is generally issued if the 24- required to evacuate the population at risk), it remains
h rainfall is expected to exceed 175 mm, or the 60-min constant regardless of the rain path of a given critical
rainfall is expected to exceed 70 mm over a substan- rainfall, Dtc1 = Dtc2 (curve B in Fig. 12).
tial part of the urban area: in such a situation, local Fig. 13 describes an operating procedure which is
radio and television stations are requested to broadcast presently being verified and tested in the study area.
the warning to the public at regular intervals. The procedure is activated after the quantitative pre-
A warning threshold could be introduced upstream cipitation rainfall forecast. Based on forecast, a pre-
to the actual triggering threshold (RBMCJ, 1985), i.e. liminary assessment is made of the probability that the
a limit which, if exceeded, activates emergency pro- warning and critical thresholds will be exceeded. The
cedures in areas susceptible to landslides triggered by forecasting phase has improved considerably in recent
rainfall. Any of the above-stated graphs that indicate years, thanks to the development of specific techni-
triggering thresholds can be used to identify warning ques such as the Doppler Radar (Smith et al., 1996),
thresholds. When identifying warning thresholds, it is which enables the accurate quantification of forecast-
important to take into account both the trend of ed rainfall.
triggering thresholds and the logistical problems that In a situation of ‘‘ordinary attention’’, in which it is
could occur during emergency evacuation procedures. assumed that the critical limits will not be exceeded,
For example, warning thresholds can be defined as the forecasting procedure is repeated. Vice versa, if
curves that are ‘‘parallel’’ to the triggering thresholds the forecast predicts rainfall exceeding the warning

Fig. 12. Criteria used to plot warning curves based on the triggering threshold. Warning curve is defined as a limit that, if exceeded, activates an
emergency procedure. Necessarily it has to be traced upstream the triggering threshold.
260 P. Aleotti / Engineering Geology 73 (2004) 247–265

Fig. 13. Operating procedure based on thresholds.

and triggering thresholds, i.e. indicating a potentially this stage, the rainfall data has to be analysed in order
critical situation, the alert phase is activated and, in to identify the beginning of critical rainfall on the
landslide-prone areas, a risk assessment procedure is time-cumulative rainfall curve, i.e. the rainfall that
initiated in real time (for the Piedmont Region a may potentially trigger soil slips. Once the point of
landslide susceptibility map is available, Aleotti et origin of the rain paths (‘‘zero point’’) has been
al., 2000). This implies the acquisition of rainfall data identified, the precipitation trend can be traced on
recorded by the involved rain gauges; data on ante- specific graphs in relation to the warning threshold
cedent rainfall for periods of 10 days prior to the that has been adopted. The alarm phase is activated
event; ‘‘historic’’ information on the rain gauges once the warning thresholds are exceeded, and the
(rainfall return times, mean annual rainfall, etc.). At emergency procedure is implemented.
P. Aleotti / Engineering Geology 73 (2004) 247–265 261

The warning system as defined in its general duration of the Yn we used Eqs. (8) and (9) to obtain
structure could be based, at least in the medium-term Eq. (10):
preliminary assessment (days), on the relationship
between antecedent rainfall and critical rainfall. A ln½NCR ¼ 8:35ln½D  10:465 ð10Þ
possible scheme is described in Fig. 14. With refer-
ence to normalised antecedent rainfall for periods of 7 which expresses the relationship between normalised
and 10 days, respectively, it has been calculated that critical rainfall (NCR) and duration (D) of the event
the curves that constitute the lower limits of 90% of (Fig. 15). This curve can be used as triggering
data are represented by Eqs. (4) and (5), respectively. threshold in Fig. 14.
Starting, for example, from the cumulative rainfall in The use of both 7 and 10 days critical triggering
n days preceding a given time t (Xn with n = 7 or 10), together is the optimal solution. In general, two
it is possible to define the volume of critical rainfall, different situations can occur: (i) rainfall antecedent
Yn, which suggests the triggering of surficial slides by to time t is mainly concentrated in the last 7 days. In
saturation. If the weather forecast states that value Yn this case Y10>Y7 and it would be more prudential to
will be reached in the upcoming hours, the warning use Y7 as critical rainfall value; (ii) rainfall is partic-
procedure is activated. In order to identify a critical ularly abundant between the 8th and 10th day ante-

Fig. 14. Scheme based on antecedent rainfall in a warning system.


262 P. Aleotti / Engineering Geology 73 (2004) 247–265

Fig. 15. Relationship between duration (D) and normalised critical rainfall (NCR).

cedent to time t. In this case, Y10 < Y7, and the warning critical event (Reichenbach et al., 1998; Crosta and
procedure must be set on the critical rainfall volume Frattini, 2001). It is necessary to: (i) identify the
indicated by Y10. event; (ii) indicate date and duration of the event;
(iii) define the area involved; (iv) identify associated
rain gauges. In addition, the following information
6. Concluding remarks must be acquired for each rain gauge: (a) geographic
coordinates; (b) daily antecedent rainfall for a period
The empirical thresholds described in this paper are of at least 15 days; (c) hourly rainfall over the whole
a fundamental element of the implemented real-time event; and (d) statistical curves for each rain gauge
warning systems. However, when using them, we and for given return periods.
must take into account several major restrictions. A Throughout this study we could not identify a
basic limitation that can be defined as conceptual significant correlation between antecedent rainfall
(Reichenbach et al., 1998) is that thresholds inevitably and critical rainfall. Nevertheless, further studies are
represent a simplification of the relationship between needed to investigate the influence of antecedent
rainfall and landslide occurrence. Rainfall is not the rainfall with the increase of the antecedent days value
direct cause of failures which, in fact, are caused by (Glade, 2000a).
the build-up of high porewater pressure in the slope, a As regards the triggering thresholds, new curves
phenomenon that is also related to hydraulic, physical have been proposed for the graphs used in the
and mechanical properties of terrain and to other literature. Comparison with other thresholds pointed
environmental factors like slope, vegetation cover, out the similarity with those proposed by Ceriani et al.
climatic characteristics of the area. A second aspect (1994) for the Italian Central Alps. The thresholds
is tied to the type and the quality of data used to could be implemented and further refined by intro-
identify triggering thresholds. In this paper we high- ducing other case studies in the analysis. The combi-
lighted the importance of accurately determining the nation of approaches presented herein with other
beginning of critical rainfall and the hour of landslide hydrological – geotechnical approaches should be
triggering, as well as the need to refer each failure to a implemented at smaller, more detailed scales, and
relatively close rain gauge. Equally important is the for specific rainfall events. An effort in this direction,
compilation of a detailed database immediately after a by applying the Montgomery and Dietrich model
P. Aleotti / Engineering Geology 73 (2004) 247–265 263

(1994), was made in north-western Piedmont in rela- of the VIII International Symposium on Landslides, Cardiff, vol.
tion to the flooding event which occurred in Vigezzo 1. Telford, London, pp. 13 – 18.
Aleotti, P., Baldelli, P., Bellardone, G., Quaranta, N., Tresso, F.,
Valley (Ossola Valley) on 7– 8 August 1978 (Campus Troisi, C., Zani, A., 2002. L’evento meteorico del 13 – 16 otto-
et al., 2001). bre 2000 nel Piemonte Settentrionale: analisi delle precipitazioni
Research has to be directed to establish the uncer- e dei processi di versante indotti. Geologia Tecnica e Ambien-
tainties associated to the critical thresholds (Chowd- tale 1, 15 – 25.
Aleotti, P., Canuti, P., Falorni, G., Fanti, R., Grimaldi, G., Guida,
hury and Flentje, 2002; Flentje and Chowdhury, 2001,
D., Lombardi, G., Pappalardo, G., Polloni, G., 2003. Assess-
2002). ment of potential debris flow inundation areas on a small allu-
The most appropriate use of critical thresholds is in vial fan in southern Italy. Proc. of the International Conference
real-time warning systems. A general outline of this on Fast Movements – prediction and prevention, Sorrento, May
procedure has been defined. A test application of this 11 – 13, 2003. In Print.
procedure has been scheduled for a real event (in back Anselmo, V., 1980. Three case studies of storm and debris flows in
north western Italy (Piemonte). Proc. of the V International Sym-
analysis) to enable its calibration and refinement. posium Interpraevent 1980, Osterr, Wasserwirtsch, Klagenfurt,
pp. 239 – 251.
Bandis, S.C., Del Monaco, G., Margottini, C., Serafini, S., Troc-
Acknowledgements ciola, A., Dutto, F., Mortara, G., 1996. Landslide phenomena
during the extreme meteorological event of 4 – 6 November
1994 in the Piedmont Region N Italy. Proc. of the VII Int.
I am grateful to Thomas Glade and Fausto Guzzetti Symposium on Landslides, Trondheim, vol. 2, pp. 623 – 628.
for having provided a very helpful review of the Bolley, S., Oliaro, P., 1999. Analisi dei debris flows in alcuni bacini
manuscript. Sincere thanks to Robin Chowdhury, campione dell’Alta Val Susa. GEAM 1999, 69 – 74 (Marzo).
Giovanni Crosta, Phil Flentje, Mario Lagorio and Bonnard, Ch., Noverraz, F., 2001. Influence of climate change
on large landslides: assessment of long term movements and
Giovanni Polloni, for discussion and valuable com-
trends. Proc. of the International Conference on Landslides:
ments during the preparation of the earlier version of causes impact and countermeasures, Gluckauf, Essen, Davos,
the paper. Thanks are owed to Jordi Corominas and to pp. 121 – 138.
an anonymous referee who greatly improved the Borga, M., Dalla Fontana, G., De Ros, Marchi, L., 1998. Shallow
paper. The Piedmont Region Authority (Direzione landslide hazard assessment using a physically based model and
Servizi Tecnici di Prevenzione, Settore Meteoidrog- digital elevation data. Environmental Geology 35, 81 – 88.
Brand, E.W., 1992. Slope instability in tropical areas. Proc. of the
rafico e Reti di Monitoraggio) is kindly acknowledged VI International Symposium on Landslides, Christchurch, vol.
for having provided the rainfall data reported in this 3. Balkema, Rotterdam, pp. 2031 – 2051.
work. The study has been partly funded by the Brand, E.W., Premchitt, J., Phillipson, H.B., 1984. Relationship
INTERREG IIC Project carried out for the Piedmont between rainfall and landslides in Hong Kong. Proc. of the
Region. IV International Symposium on Landslides, Toronto, vol. 1,
pp. 377 – 384.
Caine, N., 1980. The rainfall intensity duration control of shallow
landslides and debris flows. Geogr. Ann. 62 (1 – 2), 23 – 27.
References Campbell, R.H., 1975. Debris flow originating from soil slip during
rainstorm in southern California. Q. Enginerring Geologist 7,
Aleotti, P., Polloni, G., 2000. Fractal structure of spatial distribution 339 – 349.
of soil slips. Proc. of the IX International Conference Interprae- Campus, S., Forlati, F., Sarri, H., Scavia, C., 2001. Surficial land-
vent 2000, vol. 1. VHB, Villach, pp. 183 – 189. slide hazard assessment based on multidisciplinary studies.
Aleotti, P., Baldelli, P., Polloni, G., 1996. Landsliding and flooding Proc. of the 14th Southeast Asian Geotechnical Conference,
event triggered by heavy rains in the Tanaro basin (Italy). Proc. Hong Kong.
of the VIII International Congress Interpraevent 1996, Gar- Cancelli, A., Nova, R., 1985. Landslides in soil debris cover trig-
misch-PartenKirchen, vol. 1, pp. 435 – 446. gered by rainstorm in Valtellina (Central Alps, Italy). Proc. of
Aleotti, P., Baldelli, P., Polloni, G., 1998. Soil slips, rock-block the IV International Conference on Landslides, Tokyo, vol. 1.
slides and stream hydraulic processes caused by heavy rains: Balkema, Rotterdam, pp. 267 – 272.
their interaction and relevant hazard. Proc. of the Second Con- Cannon, S.H., 1988. Regional rainfall-threshold conditions for
ference On Environmental Management ICEM2, Wollongong, abundant debris-flow activity. In: Ellen, S.D., Wieczorek, G.F.
vol. 1. Elsevier, Amsterdam, pp. 553 – 564. (Eds.), Landslides, Floods and Marine Effects of the Storm of
Aleotti, P., Baldelli, P., Govi, M., Polloni, G., Villani, B., 2000. 3 – 5 January, 1982 in the San Francisco Bay Region, California,
Landslide hazard assessment in the Po River Basin (Italy). Proc. pp. 35 – 42. USGS Prof. Paper 1434.
264 P. Aleotti / Engineering Geology 73 (2004) 247–265

Cannon, S.H., Ellen, S.D., 1985. Rainfall conditions for abundant Gardland, G.G., Olivier, M.J., 1993. Predicting landslides from
debris avalanches, San Francisco Bay region, California. Cali- rainfall in a humid, subtropical region. Geomorphology 8,
fornia Geology 38 (12), 267 – 272. 165 – 173.
Ceriani, M., Lauzi, S., Padovan, N., 1994. Rainfall thresholds trig- Glade, T., 2000a. Modelling landslide triggering rainfall thresholds
gering debris flows in the alpine area of Lombardia Region at a range of complexities. Proc of the VIII International
Central Alps – Italy. I Convegno internazionale per la protezione Symposium on Landslides, Cardiff, vol. 2. Telford, London,
e lo sviluppo dell’ambiente montano ‘‘Man and Mountain’’94’’, pp. 633 – 640.
pp. 123 – 139. Glade, T., 2000b. Modelling landslide-triggering rainfalls in differ-
Chiarle, M., Luino, F., 1996. Colate detritiche torrentizie sul Monte ent regions of New Zealand—the soil water status model. Zeits-
Mottarone innescate dal nubifragio dell’8 luglio 1996 (Lago chrift für Geomorphologie N.E. 122, 63 – 84.
d’Orta, Piemonte). Atti del Convegno Internazionale ‘‘La pre- Glade, T., Crozier, M., Smith, P., 2000. Applying probability deter-
venzione delle catastofi idrogeologiche: il contributo della mination to refine landslide-triggering rainfall thresholds using
ricerca scientifica’’, Alba (Italy), vol. 2, pp. 231 – 245. empirical ‘‘antecedent daily rainfall model’’. Pure and Applied
Chowdhury, R.N., Flentje, P., 2002. Uncertainties in rainfall-in- Geophysics 157, 1059 – 1079.
duced landslide hazard. Quarterly Journal of Engineering Geol- Govi, M., Mortara, G., Sorzana, P., 1985. Eventi idrologici e frane.
ogy and Hydrogeology 35 (Part 1), 61 – 71. Geologia Applicata e Idrogeologia 20 (2), 395 – 401.
Clarizia, M., Gullà, G., Sorbino, G., 1996. Sui meccanismi di Guidicini, G., Iwasa, O.Y., 1977. Tentative correlation between
innesco dei soil slip. Atti del Convegno Internazionale ‘‘La rainfall and landslides in a humid tropical environment. Bulletin
prevenzione delle catastofi idrogeologiche: il contributo della of the International Association of Engineering Geology 16,
ricerca scientifica’’. Alba (Italy), vol. 1, pp. 585 – 597. 13 – 20.
Corominas, J., 2000. Landslide an climate. Proc of the VIII Inter- Guzzetti, F., Crosta, G., Marchetti, M., Reichenbach, P., 1992. De-
national Symposium on Landslides, Cardiff. Keynote Paper bris flows triggered by the July, 17 – 19, 1987 storm in the
Published in Electronic Form. Valtellina Area (Northern Italy). Proc. of the VII International
Corominas, J., Moya, J., 1999. Reconstructing recent landslide ac- Congress Interpraevent 1992, Bern, vol. 2, pp. 193 – 204.
tivity in relation to rainfall in the Llobregat river basin, Eastern Iiritano, G., Versace, P., Sirangelo, B., 1998. Real time estimation of
Pyrenees, Spain. Geomorphology 30, 79 – 93. hazard for landslides triggered by rainfall. Environmental Geol-
Crosta, G., 1998. Regionalization of rainfall thresholds: an aid to ogy 35 (2 – 3), 175 – 183.
landslide hazard evaluation. Environmental Geology 35 (2 – 3), Jibson, R.W., 1989. Debris Flow in Southern Porto Rico. In:
131 – 145. Schultz, Jibson (Eds.), Landslide processes of the Eastern Unit-
Crosta, G., Frattini, P., 2001. Rainfall thresholds for triggering soil ed States and Puerto Rico. Geological Society of America Spe-
slips and debris flow. Proc. Of EGS 2nd Plinius Conference cial Paper, vol. 236, pp. 29 – 55.
2000, Mediterranean Storms, Siena, pp. 463 – 488. Keefer, D.K., Wilson, R.C., Mark, R.K., Brabb, E.E., Brown, W.M.,
Crosta, G.B., Dal Negro, P., Frattini, P., 2003. Distributed model- Ellen, S.D., Harp, E.L., Wieczorek, G.F., Alger, C.S., Zatkin,
ling of shallow landsliding in volcanoclastic soils. Engineering R.S., 1987. Real time landslide warning system during heavy
Geology (this volume). rainfall. Science 238, 921 – 925.
Crozier, M.J., 1999. Prediction of rainfall triggered landslides: a test Luino, F., Ramasco, M., Susella, G., 1994. Atlante dei Centri abitati
of the antecedent water status model. Earth Surface Processes instabili piemontesi. Gruppo Nazionale per la Difesa dalle Cat-
and Landforms 24, 825 – 833. astrofi Idrogeologiche, programma speciale: Studio Centri Abi-
Crozier, M.J., Glade, T., 1999. Frequency and magnitude of land- tati Instabili, pubblicazione 964. 245 pp.
sliding: fundamental research issues. Zeitschrift für Geomorpho- Lumb, P., 1975. Slope failure in Hong Kong. Quarterly Journal
logie N.F. 15, 141 – 155. Engineeering Geologist 8, 31 – 65.
Flentje, P., Chowdhury, R., 2001. Aspects of Risk Management for Montgomery, D.R., Dietrich, W.E., 1994. A physically-based model
Rainfall-Triggered Landsliding. Proc. Symposium on Engineer- for the topographic control on shallow landsliding. Water
ing and Development in Hazardous Terrain, New Zealand Geo- Resources Research 30, 1153 – 1171.
technical, pp. 143 – 150. Morgan, B.A., Wieczorek, G.F., Campbell, R.H., Gori, P.L., 1997.
Flentje, P., Chowdhury, R.N., 2002. Landsliding in an Urban Area Debris flow hazards in areas affected by the June 27, 1995 storm
‘‘Photographic Feature’’. Quarterly Journal of Engineering Ge- in Madison County, Virginia. USGS Open File Report, 97 – 438.
ology and Hydrogeology 35 (Part 1), 5 – 9. Moser, M., Hohensinn, F., 1983. Geotechnical aspect of soil slips in
Flentje, P., Chowdhury, R.N., Tobin, P., 2000. Management of land- Alpine regions. Engineering Geolologist 19, 185 – 211.
slides triggered by a major storm event in Wollongong, Aus- Onodera, T., Yoshinaka, R., Kazama, H., 1974. Slope failures
tralia. Proc. of the II International Conference on Debris-Flow caused by heavy rainfall in Japan. Proc. of the II International
Hazards Mitigation, Mechanics, Prediction and Assessment, Congress International Association of Engineering Geology,
Taipei, pp. 479 – 487. Sao Paulo, Brasil, vol. 11, pp. 1 – 10.
Forlati, F., Lancellotta, R., Osella, A., Scavia, C., Veniale, F., 1995. Paronuzzi, P., Del Fabbro, M., Maddaleni, P., 2002. Frane super-
The role of swelling marls in planar slides in the Langhe Re- ficiali tipo slide debris flow causate dal nubifragio del 21/22
gion. Proc. VII International Symposium on Landslides, Trond- giugno 1996 nella Val Chiarsò (Alpi Carniche, Friuli). Memorie
heim, vol. 2. Balkema, Rotterdam, pp. 721 – 725. della Societa Geologica Italiana 57, 443 – 452.
P. Aleotti / Engineering Geology 73 (2004) 247–265 265

Polloni, G., Ceriani, M., Lauzi, S., Padovan, N., Crosta, G., from upslope thunderstorm in the Central Appalachians: the
1992. Rainfall and soil slipping events in Valtellina. Proc. Rapidan storm of June 27, 1995. Water Resources Research
of the VI International Symposium on Landslides. Christchurch 32 (10), 3099 – 3113.
1, 183 – 188. Sutera Sardo, P., Beretta, G.P., De Luca, D.A., Masciocco, L., 1996.
Polloni, G., Aleotti, P., Baldelli, P., Nosetto, A., Casavecchia, K., Frane superficiali innescatesi nei dintorni di Ceva (CN) in con-
1996. Heavy rain triggered landslides in the Alba area during seguenza dell’evento meteorologico del Novembre 1994. Atti
November 1994 flooding event in the Piemonte Region (Italy). del Convegno Internazionale ‘‘La prevenzione delle catastofi
Proc. VII International Symposium on Landslides, Trondheim, idrogeologiche: il contributo della ricerca scientifica’’, Alba
vol. 3, pp. 1955 – 1960. (Italy), vol. 1, pp. 575 – 583.
Premchitt, J., 1997. Warning system based on 24-hour rainfall in Terlien, M.T.J., 1998. The determination of statistical and determin-
Hong Kong. Manual for zonation on areas susceptible to rain- istic hydrological landslide-triggering thresholds. Environmen-
induced slope failure. Asian Technical Committee on Geotech- tal Geology 35 (2 – 3), 125 – 130.
nology for Natural Hazards in International Society of Soil Me- Wieczorek, G.F., 1987. Effect of rainfall intensity and duration
chanics and Foundation Engineering, pp. 72 – 81. on debris flows in central Santa Cruz Mountains, Califor-
RBMCJ, 1985. Study on the Disaster Prevention System to Mitigate nia. In: Costa, Wieczorek (Eds.), Debris Flows/Avalanches:
Sediment Disasters Caused by Heavy Rains, Overall Debris Processes, Recognition and Mitigation. Reviews in Engi-
Countermeasures, River Bureau, Ministry of Construction of neering Geology, vol. 7. Geological Society of America,
Japan. pp. 23 – 104.
Regione Piemonte-Direzione Servizi Tecnici di Prevenzione, 1998. Wieczorek, G.F., 1996. Landslide triggering mechanisms. In: Turner,
Eventi alluvionali in Piemonte: 2 – 6 novembre 1994, 8 luglio Shuster (Eds.), Landslides: Investigation and Mitigation. Trans-
1996, 7 – 10 ottobre 1996. Torino 1998. 415 pp. portation Research Board-National Research Council, Special
Regione Piemonte – Direzione Servizi Tecnici di Prevenzione, Report, vol. 247, pp. 76 – 90.
2000. Eventi alluvionali in Piemonte: 10 – 14 giugno 2000. Tor- Wieczorek, G.F., Guzzetti, F., 1999. A review of rainfall thresholds
ino 2000. 120 pp. for triggering landslides. Proc. of the EGS Plinius Conference,
Reichenbach, P., Cardinali, M., De Vita, P., Guzzetti, F., 1998. Maratea, Italy October 1999, pp. 407 – 414.
Regional hydrological thresholds for landslides and floods in Wieczorek, G.F., Morgan, B.A., Campbell, R.H., 2000. Debris flow
the Tiber River Basin (central Italy). Environmental Geology hazards in the Blue Ridge of Central Virginia. Environmetal and
35 (2 – 3), 146 – 159. Engineering Geoscience VI (1), 3 – 23.
Sandersen, F., Bakkeoi, S., Hestnes, E., Lied, K., 1996. The influ- Wilson, R.C., Wieczorek, G.F., 1995. Rainfall thresholds for the
ence of meteorological factors on the initiation of debris flows, initiation of debris flow at La Honda, California. Environmental
rockfalls, rockslides and rockmass stability. Proc. of the VII and Engineering Geoscience 1 (1), 11 – 27.
International Symposium on Landslides, Trondheim, vol. 1. Bal- Wilson, R.C., Mark, R.K., Barbato, G.E., 1992. Operation of real-
kema, Rotterdam, pp. 97 – 114. time warning system for debris flows in the San Francisco Bay
Sirangelo, B., Braca, G., 2001. L’individuazione delle condizioni di area, California. In: Shen, H.W., Wen, F. (Eds.), Hydraulic En-
pericolo di innesco delle colate rapide di fango. Applicazione gineering ’93. Proceedings of the 1993 Conference, Hydraulics
del modello FlaIR al caso di Sarno. Atti del Convegno: ‘‘Il Division, 1993, vol. 2. American Society of Civil Engineers,
dissesto idrogeologico: inventario e prospettive’’, Roma. San Francisco, CA, pp. 1908 – 1913.
Smith, J.A., Baeck, M.L., Steiner, M., 1996. Catastrophic rainfall

You might also like