Advances in Geosciences, 2, 301–304, 2005

SRef-ID: 1680-7359/adgeo/2005-2-301 Advances in

European Geosciences Union Geosciences
© 2005 Author(s). This work is licensed
under a Creative Commons License.

Analysis of infiltration, seepage processes and slope instability

mechanisms during the November 2000 storm event in Tuscany
V. Tofani, S. Dapporto, P. Vannocci, and N. Casagli
Earth Sciences Department, University of Florence, Italy
Received: 24 October 2004 – Revised: 10 February 2005 – Accepted: 16 February 2005 – Published: 15 September 2005

Abstract. On the days 20–21 November 2000, a storm of tors such as lithology, morphology and land cover, influence
exceptional intensity triggered over 50 landslides within the the type of landslide. Deep-seated landslides are often trig-
province of Pistoia in Tuscany (Italy). These failures are gered by moderate intensity rainfall distributed over long pe-
mostly of complex type, originating as rotational or transla- riods (Corominas, 2001) whereas superficial landslides such
tional landslides, and transforming into flows. Two of these as soil slips and debris flows are triggered by short duration,
landslides were investigated in this paper by modelling the intense precipitation. The response of the material involved
ground water infiltration process, the pore water pressure is largely dependent on its permeability. Clayey materials,
variations, both positive and negative, and the effects of these for example, are more sensitive to long, moderate intensity
variations on slope stability during the rainfall event. Mor- rainfall, as opposed to sandy soils, that tend to respond to
phometric and geotechnical analyses were carried out for intense short duration precipitation.
both sites through a series of in-situ and laboratory tests, the The critical quantity of precipitation defined as “the
results of which were used as input for the modelling process. amount of rainfall from the time (“zero point”) in which a
In a first step the surface infiltration rate was simulated us- sharp increase in rainfall intensity is observed and the trigger-
ing a modified Chu (1978) approach for the Green and Ampt ing of the (first) landslide” (Aleotti, 2004) is related to slope
(1911) equations in case of unsteady rainfall together with a morphology, to the mechanical and hydrological properties
surficial water balance. A finite element seepage analysis for of the soil and to the vegetation cover. Statistical analyses of
transient conditions was then employed to model the changes landslide characteristics and of the corresponding triggering
in pore water pressure during the event, using the computed rainfall characteristics have led to the formulation of various
infiltration rate as the ground surface boundary condition. Fi- methods for assessing landslide-triggering rainfall thresholds
nally, once again using the data from the previous step as (e.g. Govi and Sorzana, 1980; Crozier, 1986; Kim et al.,
input, the limit equilibrium Morgenstern-Price (1965) slope 1991; Glade et al., 2000; Sidle and Dhakal, 2002; Ko Ko
stability method was applied to calculate the variations in the et al., 2004).
factor of safety during the event and thereby determine the In this work the mechanism behind rainfall-triggered land-
most critical time of instability. In both sites this method slides is modelled by using combined hydrological and sta-
produced a curve for the factor of safety that indicated that bility analyses. This method is based on the different terrain
the most critical time for failure occurred a few hours after response according to its geological, physical, hydrogeologi-
the peak of rainfall. cal and mechanical characteristics. In addition, in sites where
morphological and anthropogenic factors influence the su-
perficial and subsurface fluctuations of the water flow, and
1 Introduction hence the stability of the slope (Terlien, 1998), the method
can be modified to take into account these factors. The main
Rainfall is considered to be the most important factor in trig- objective is to investigate the instability mechanisms behind
gering slope failure. During rainfall, water infiltrates into rainfall-triggered landslide events, based on detailed mod-
the ground and increases the pore water pressure, weak- elling of infiltration and seepage processes during the storm
ening the soil and facilitating landsliding. The frequency of 20–21 November 2000. More specifically, the work has
and magnitude of rainfall events, together with other fac- been conducted with the aim of increasing the understanding
between the variations in positive and negative pore water
Correspondence to: V. Tofani pressure during the rainfall event and their effect on slope
([email protected]) stability (Dapporto et al., 2002).
302 V. Tofani et al.: Analysis of infiltration, seepage processes and slope instability mechanisms

ITALY
study
area

Rome

Fig. 1. Map of the study area with locations of landslides (1), nearby towns (2) and raingauge stations (3).

Table 1. Results of in-situ tests for both landslide sites. Gumbel distribution method to the available last 30 years
rainfall record. The storm triggered over 50 landslides, in-
Site φ c φb ua -uw ks volving an area of approximately 50 km2 .
◦ kPa ◦ kPa m/s The two landslides selected for investigation occurred
near Uzzano (site 1) and Massa e Cozzile (site 2) (Fig. 1).
1. Uzzano 31.8 5.7 31.8 8.4 1.1E-6 Both landslides are classified as complex slope movements
2. Cozzile 35.7 4.7 31.7 7.6 1.3E-6
(Hutchinson, 1988), composed of a rotational-translational
failure that evolves into a flow (Fig. 2).

2 Study area and rainfall event 3 Morphometry and characterisation of soil properties

The predominant rock type in the study area is a flysch for- The topography of the two landslide sites was determined
mation (Macigno), composed of quartz and feldspar sand- by means of a differential GPS survey of the pre- and post-
stone alternated with layers of siltstone. In the mid and up- failure slope profiles. The pre-failure profile was obtained by
per sections of the valley, where most landslides occurred, averaging the undisturbed profiles surveyed on both sides of
the stratigraphy consists of a 1.5 to 2 m thick layer of collu- the landslide body following the rainfall event.
vial soil overlying the bedrock. Soil geotechnical parameters were determined from a se-
The local topography is characterised by hills with mod- ries of in-situ and laboratory tests. The in-situ tests included
erate to high slope gradients and elevations ranging from the Borehole Shear Test (Lutenegger and Halberg, 1981;
20 m a.s.l. to about 900 m a.s.l. Dapporto et al., 2000), matric suction measurements with
Annual rainfall in this area is characterised by two peaks, a tensiometer, and a constant head permeameter test per-
one in November and another in March, with a dry summer formed with an Amoozemeter (Amoozegar, 1989) (Table 1).
during which the mean temperature is over 22◦ C. Mean an- Grain size distribution, Atterberg limits, the phase relation-
nual rainfall increases with relief, ranging from 1100 mm in ship analysis and direct shear tests were conducted in the lab-
the valley bottom to about 1800 mm at the higher elevations. oratory (Table 2).
On 20–21 November 2000, a cyclonic storm precipitated
about 200 to 220 mm of rainfall over a period of 38–40 h 4 Infiltration analysis
in the Montecatini Terme area (Province of Pistoia), with a
maximum intensity of 17 mm/h. The event has an estimated The infiltration rate at the surface during the rainstorm was
return period of over 100 years, determined by applying the predicted by using a modified Chu (1978) approach for the
V. Tofani et al.: Analysis of infiltration, seepage processes and slope instability mechanisms 303

Table 2. Results of laboratory tests for both landslide sites.

Site GF SF MF CF LL PL γsat γd Gs n 
φcv φr c’
% % % % % % kN/m3 kN/m3 – % ◦ kPa
1. Uzzano 12 43 35 10 26 16 19.7 15.8 2.68 40 32 – 0.5
2. Cozzile 1 85 13 1 29 – 18.9 14.7 2.67 44 35 18 0

Green and Ampt (1911) equations in case of unsteady rainfall

combined with a water balance at the surface. A B
The rainfall recorded at each of the rainfall gauges nearest
the sites was discretised into a series of one-hour time steps
with constant rainfall rates (Fig. 1). The infiltration rate and
the start and end times of ponding conditions were computed
at the end of each time step.
The data required for the infiltration analysis includes:
the saturated permeability (ks ); the volumetric water con-
tent before and after wetting (M=θ s−θ i) and the matric suc-
tion at the wetting-front (S). The matric suction measured in
the field in post-landslide conditions (8.4 kPa at site 1 and Fig. 2. Study sites. A: Site 1, Uzzano landslide, B: Site 2, Massa
7.6 kPa at site 2) was modified to replicate the presumed ini- and Cozzile landslide.
tial pore water pressure in pre-landslide conditions, in which
heavy antecedent rainfalls were recorded during the 20 days
prior to the storm. For this reason a matric suction value 6 Stability analysis
of 5 kPa was used. The volumetric water content functions,
before and after wetting, were obtained from the literature The stability analysis used the positive and negative pore wa-
for soils with a similar grain size distribution while the sat- ter pressure distributions obtained from the seepage analysis
urated water content (θs ) was derived from the porosity (n) as input at each time step. SLOPE/W v.4 (Geo-Slope Interna-
measured in the laboratory. In both sites, at the beginning of tional Ltd.), a software that applies limit equilibrium meth-
the storm the modelled infiltration rate is equal to the rain- ods, was used to model slope stability. The Morgenstern-
fall rate. Subsequently, following the onset of ponding con- Price (1965) method was selected to compute the factor of
ditions, the infiltration rate becomes equal to the maximum safety of the soil cover for each of the time steps used in the
infiltration rate for those soil conditions or, in other words, to finite element seepage analysis. The failure surface was de-
the infiltration capacity. The infiltration rate asymptotically termined from the GPS measurements and was maintained
tends to the value of the saturated permeability (Ks ). constant during the entire event.

5 Seepage analysis 7 Results

Unsaturated and saturated flow within the slopes was mod- The results of the combined infiltration-seepage-stability
elled with a finite element seepage analysis using SEEP/W analysis indicate a similar behaviour of the Uzzano and
v.4 (Geo-Slope International Ltd.) software. The slope sec- Massa e Cozzile landslides. In both sites the factor of safety
tions were divided into a mesh of quadrilateral elements, starts to decrease after the start of the rainfall (Figs. 3 and 4).
1152 for Uzzano and 456 for Massa e Cozzile. The results This is due to the reduction of matric suction produced by the
from the infiltration analysis were used as input data for the water that infiltrates into the soil. As the event progresses,
seepage analysis by assigning the computed infiltration rate the wet front descends through the soil until it reaches the
versus time function to the ground surface nodes. Zero flux impermeable bedrock, at which point a saturated layer be-
conditions were assigned to the left, right and lower bound- gins to build up. During this phase in both sites the factor
aries. An initial matric suction value of 5 kPa was used. The of safety decreases rapidly. In site 1 the factor of safety falls
hydraulic properties of the materials (characteristic and k- below 1 at step 19 (Fig. 3) while in site 2 this occurs at step
curves) were estimated from the grain-size distribution of 13 (Fig. 4), respectively 32 and 24 h after the start of the rain-
the matrix. Finally, the simulation was performed dividing fall. The factor of safety then slowly begins to increase again
the rainfall event into variable length time steps in order to as the pore water pressure diminishes following the end of
replicate as closely as possible the complexity of the rainfall rainfall, and stability conditions are reached four and three
event. days after the beginning of the event, respectively.
304 V. Tofani et al.: Analysis of infiltration, seepage processes and slope instability mechanisms

2.0 20 References
1.8
1 18

2 16 Aleotti, P.: A warning system for rainfall-induced shallow failures,

1.6 14 Engineering Geology, 73, 247–265, 2004.
Factor of safety

Rainfall (mm)
12 Amoozegar, A.: Compact constant head permeameter for measur-
1.4
10
1.2
ing saturated hydraulic conductivity of the vadose zone, Soil Sci-
8
ence Society of American Journal, 53, 1356–1361, 1989.
1.0 6
Chu, S. T.: Infiltration during an unsteady rain, Water Resources
4
0.8
2
Research, 14, 461–466, 1978.
0.6 0
Corominas, J.: Landslides and Climate, Keynote Lectures from the
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 8th International Symposium on Landslides, No. 4, 1–33, 2001.
Time (hours)
Crozier, M. J.: Landslides: Causes, Consequences and Enviror-
ment, Croom Helm, London, pp. 252, 1986.
Fig. 3. Site 1. Uzzano. Trend of the computed safety factor and Dapporto, S., Focardi, P., and Vannocci, P.: La parametrizzazione
rainfall data during the 20-21 November 2000 event (1- factor of geotecnica di un materiale non saturo mediante prove in sito, Ge-
safety, 2- rainfall). ologia Tecnica & Ambientale, 3, 5–11, 2000.
Dapporto, S., Falorni, G., Tofani, V., and Vannocci, P.: Anal-
2.0 20 isi delle variazioni di pressione interstiziale alla base dei movi-
1.8 1 18 menti franosi durante l’evento del novembre 2000 in Toscana,
2 16 Conservazione dell’ambiente e rischio idrogeologico, Convegno
1.6 14 nazionale Assisi, Italia, 11–12 dicembre 2002.
Factor of safety

Rainfall (mm)

12
1.4 Glade, T., Crozier, M., and Smith, P.: Applying Probability Deter-
10
1.2
mination to Refine Landslide-triggering Rainfall Thresholds Us-
8
ing an Empirical “Antecedent Daily Rainfall Model”, Pure Appl.
1.0 6

4
Geophys., 157, 1059–1079, 2000.
0.8
2
Govi, M. and Sorzana, P. F.: Landslides susceptibility as a function
0.6 0 of critical rainfall amount in Piedmont Basins, Studia Geomor-
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
phologica Carpatho-Balcanica, 14, 43–61, 1980.
Time (hours)
Green, W. H. and Ampt, G. A.: Studies on soil physics: 1. Flow of
air and water through soils, J. Agr. Sci., 4, 1–24, 1911.
Fig. 4. Site 2. Massa e Cozzile. Trend of the computed safety factor Hutchinson, J. N.: General report: Morphological and geotechnical
and rainfall data during the 20–21 November 2000 event (1- factor
parameters of landslides in relation to geology and hydrogeology,
of safety, 2- rainfall).
Proc. Fifth International Symposium on Landslides, Rotterdam,
Balkema, 1, 3–36, 1988.
Kim, S. K., Hong, W. P., and Kim Y. M.: Prediction of rainfall-
8 Conclusions triggered landslides in Korea, in: Proceedings of the 6th In-
ternational Symposium on Landslides, edited by: Bell, D. H.,
The results indicate that the simulation of satu- Christchurch, New Zealand, Balkema, Rotterdam, 989–994,
rated/unsaturated flow within the soil with a finite element 1991.
seepage analysis, coupled with an infiltration model to Ko Ko, C., Flentje, P., and Chowdhury, R.: Interpretation of prob-
determine the actual infiltration rate at the surface, can ability of landsliding triggered by rainfall, Landslides, 1, 263–
provide useful insight for the reconstruction of pore water 275, 2004.
pressure variations, total head and volumetric water content Lutenegger, A. J. and Halberg, G. R.: Borehole Shear Test in
in response to variable intensity rainfall. The assessment of geotechnical investigation, Special Technical Publ. American
pore pressure changes at all nodes along the potential failure Society for Testing and Materials, 740, 566–578, 1981.
surface also allows an accurate slope stability analysis to be Morgenstern, N. R. and Price, V. E.: The analysis of the stability of
performed. The trend of the factor of safety indicates that general slip surfaces, Geotechnique, 15, 79–93, 1965.
slope failure is likely to occur a few hours after the peak of Sidle, R. C. and Dhakal, A. S.: Potential effects of environmental
the rainfall, and that in both sites slope instability conditions change on landslide hazards in forest environments, in: Environ-
persist for the following 4 days. mental Change and Geomorphic Hazards in Forests, edited by:
Sidle, R. C., IUFRO Research Series, CAB International Press,
Edited by: L. Ferraris
Oxen, UK, No. 9, pp. 123–165, 2002.
Reviewed by: anonymous referees
Terlien, M. T. J.: The determination of statistical and deterministic
hydrological landslide-triggering thresholds. Environmental ge-
ology, 35, 2–3 August 1998.