Professional Documents
Culture Documents
com
Abstract
At the end of March 2006, the Czech Republic (CZ) witnessed a fast thawing of an unusually thick snow cover in conjunction with massive
rainfall. Most watercourses suffered floods, and more than 90 shallow landslides occurred in the Moravian region of Eastern CZ, primarily in nonforested areas. This region, geologically part of the Outer Western Carpathians, is prone to landslides because the bedrock is highly erodible
Mesozoic and Tertiary flysch.
The available meteorological data (depth of snow, water equivalent of the snow, cumulative rainfall, air and soil temperatures) from five local
weather stations were used to construct indices quantitatively describing the snow thaw. Among these, the Total Cumulative Precipitation (TCP)
combines the amount of water from both thawing snow and rainfall. This concurrence of rain and runoff from snow melt was the decisive factor in
triggering the landslides in the spring.
The TCP index was applied to data of snow thaw periods for the last 20 years, when no landslides were recorded. This was to establish the safe
threshold of TCP without landslides. The calculated safe threshold value for the region is ca. 100 mm of water delivered to the soil during the
spring thaw (corresponding to ca. 11 mm day 1). In 2006, 10% of the landslides occurred under or at 100 mm of TCP. The upper value of 155 mm
covered all of the landslides.
2007 Elsevier B.V. All rights reserved.
Keywords: Shallow landslides; Antecedent rainfall; Snow thaw; Landslide threshold; Outer Western Carpathians
1. Introduction
Shallow landslides in the Outer Western Carpathians are
almost always associated with extreme rainfall (e.g. Gil, 1997;
Krej et al., 2002; Kudrna et al., 2003). In the spring of 2006,
the thawing of the thick snow cover occurred together with
massive rainfall. This article attempts to uncover the causes of
the resulting landslide disaster, and to establish the threshold
amount of water from snowmelt and rainfall that when soaked
into the soil, will trigger shallow landslides.
For evaluating the rainfall threshold for landsliding, various
modifications of the Antecedent Rainfall Index (ARI) and the
Antecedent Daily Rainfall Model have been used (e.g. Caine,
1980; Glade et al., 2000; Godt et al., 2006). These usually
consider only liquid precipitation. However, the situation under
Corresponding author. Tel.: +420 585 634 177; fax: +420 585 634 002.
E-mail address: michal.bil@upol.cz (M. Bl).
0169-555X/$ - see front matter 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.geomorph.2007.11.004
study here is unique in that the melt of the snow cover and the
liquid rainfall are combined. The snow cover can be regarded as
a reservoir of water with a known content. As the depth of snow
and its water equivalent are measured on a regular basis, these
data can be used for calculating the total amount of water
delivered into the soil. Gerstel et al. (1997) discussed extremely
fast snow thawing accompanied by rainfall, followed by
landslides in Washington State, USA.
In ascertaining the rainfall threshold for landsliding, most methods take the known dates of landslide occurrences as their input
parameter. Usually several episodes with a substantial number of
landslides are considered. Although these relatively simple models
of water balance are not substantiated physically, they are still
feasible for initial approximation. The same is true for various other
indices describing landslide occurrence such as the Cumulative
Event Rainfall (Corominas and Moya, 1999). Most of the work on
rainfall threshold determination has been done in Italy (Govi and
Sorzana, 1980; Reichenbach et al., 1998; Calcaterra et al., 2000;
247
Fig. 1. Study area. Black dots landslides in the spring of 2006, flags weather stations, LU Luhaovice, VI Vizovice, HL Horn Lhota, VK Valask
Klobouky, ST ttn.
248
Table 1
Landslides with known dates of origin
Date
3/27
3/28
3/29
3/30
3/31
4/1
4/2
4/3
4/4
Sum
Frequency
28
of the study area, snow thaw occurs within several days. As the
point of reference we took the last day when a continuous snow
cover was recorded, since small patches of snow can survive in
shaded locations for a long time.
We employed the following three indices.
Snow thaw period, which begins when the average daily
temperature rises above 1 C and ends seven days after the
loss of a continuous snow cover. The beginning is thus
located within days when the day-time temperature is above
0 C and freezing occurs overnight. The average daily temperature is calculated as the sum of temperatures at 7 a.m.,
2 p.m. and twice the temperature at 9 p.m. divided by four.
This is the standard procedure used by the Czech Hydrometeorological Institute, CHMI. The end of the snow thaw
period is not defined as the last day of continuous snow cover
but extra seven days are added because snow can remain in
isolated patches, and runoff water stays in pools on the soil
surface, gradually infiltrating into the ground.
Snow Thaw Rate (STR) is the average decrement (in mm
day 1) of the snow water equivalent (SWE) from the beginning of the snow thaw period until the last day of continuous snow cover. Since the logs of SWE are available only
once a week, the exact value at the beginning of the snow
thaw period may not be known, and STR suffers from this
uncertainty. In addition, an auxiliary index of the thawing
rate was calculated from the snow depth (SD) (in cm day 1).
However, the latter index is not precise because of the
settling or compaction of snow with time.
Fig. 2. Cumulative rainfall (mm) and snow depth (SD; cm) during the 2006 landslide episode at the station of Luhaovice, March 20April 5.
249
Fig. 3. Pentades of rainfall, snow depth (SD), Total Cumulative Precipitation (TCP) and change in SD at the Vizovice station.
250
Table 2
Station Luhaovice weather conditions during Snow Thaw Periods 19882006
Year
88
89
90
91
92
93
94
95
12
16.9
8
17.0
4.3
32.7
2.7
15
8.2
11
14.0
2.3
17.5
1.2
10
21.3
5
10.5
10.5
31.8
3.2
10
0.3
11
22.4
5.0
22.7
1.9
20
21.2
12
44.1
4.0
62.4
3.1
14
26.0
17
23.0
11.5
48.1
3.4
11
21.2
22
31.2
7.8
52.1
4.3
13 12
25.2 8.9
13
5
26.2 12.1
6.6 1.7
37.1 21.4
2.9 1.4
96
97
98
15 13
20.7 6.1
16
6
27.7 9.0
6.9 3.0
48.4 15.1
3.2 1.2
99
00
01
02
14
39.0
20
33.3
11.1
65.4
4.7
20
38.5
22
44.9
4.5
64.7
3.2
12
25.4
13
16.0
3.2
41.4
3.2
17 12
19.0 0.3
23
6
45.4 13.5
5.7 1.9
55.4 13.8
3.3 0.9
03
04
05
15
6.8
15
33.1
5.5
37.2
2.5
20
16
6.7 65.1
36
33
93.1 78.4
7.8
9.8
99.8 143.5
5.0
9.0
06
Daily average TCP is defined as TCP divided by number of days since the record of SWE to the end of STP.
rejection for the data until 2006. This conclusion was also validated
by the so-called omnibus test (D'Agostino et al., 1990) that uses
both empirical moments simultaneously.
Second, the Kolmogorov goodness-of-fit test was performed.
It takes better account of tails of a frequency distribution and is
also recommended for use with small sample sizes (Lehmann,
1999, p.341). The maximum difference between the empirical
and theoretical (Gaussian) distribution functions (with the
parameters and 2 estimated from the data) for records including 2006 is 0.185, leading to non-rejection of normality on
0.05-level (the non-rejection being by a narrow margin, though).
The results of all the above-performed tests do not contradict the
interpretation that TCP fits the normal model and the value of
2006 deviates from the model in the sense that it constitutes a
rare observation originating from the tail of the distribution.
The TCP values up to 2005 show no linear increasing or
linear decreasing time trend (t = 1.63 in a straight-line fit) and
thus only a constant value of the average may be used to express
a trend. The sign test on residuals from the average did not detect
any systematic error (the test statistic 0.04). Therefore the
residuals can be considered random and the model of constant
trend is applicable. The 95-percent confidence interval for the
mean value, calculated from data until 2006 under normal
model, is [32.69; 63.15].
Assuming a normal distribution N(, 2), we can estimate
the probability of the occurrence of the extreme TCP value in
2006. Estimating the mean as 47.92 and the standard deviation 2 as 998.31, we get P(X 143.5) = 0.0012, that is, the
chance of observing an equal or higher TCP value is 0.12%.
Table 3
Station Vizovice weather conditions during Snow Thaw Periods 19862006
Year
86
87
88
89
90
91
92
93
94
95
96
97
98
99
00
01
02
03
04
05
12
14
11
10
10
10
13
14
16
13
15
15
10
12
13
15
11
12
18
1.9 4.4 11.5 7.6 7.6 15.9 6.5 1.2 20.0 16.0 2.0
9.6 19.2 49.7 50.6 50.0 45.8 54.0 16.0 57.0 59.8 28.0
2.4
5.3
2.7
1.7
3.9
0.7
1.5
3.6
3.2
4.2
3.1
3.6
1.1
4.8
4.6 2.2
9.7
67.8
4.5
06
16
0.7
2.6
5.2
10.2
Luhaovice
TCP
[mm]
Vizovice
TCP
[mm]
Luhaovice
SWE
[mm]
Luhaovice
Rainfall
[mm]
Average
until
2005
Average
until
2006
Median Median
until
until
2005
2006
Std.
deviation
until 2005
Std.
deviation
until 2006
42.61
47.92
39.3
41.4
22.13
31.60
42.06
47.82
42.4
45.8
22.56
34.37
28.69
31.31
24.6
26.2
19.99
22.53
17.32
19.83
19.85
20.7
11.54
15.58
251
P X zk X n P X k X n z0
1 U k 1A= r2 1 k 2 =n ;
where k is a given constant (N 0), is the standard normal
distribution function and all the observations are assumed
independent. After substituting the estimated values for and 2
and assuming k = 3.4 (in 2006 the value of TCP was 3.4 times
higher than the antecedent average), we get a probability of 0.0024
for exceeding the k-multiple of the past mean value of TCP.
Similarly, for k = 3.9 this probability in Vizovice is 0.0012. Thus,
both stations are on the order of magnitude 10 3.
Let us now examine SWE and rainfall in 2006 in Luhaovice.
The probability of observing a rainfall of 56.1 mm or higher is
0.0020, and the probability of exceeding the 3.8-multiple of the
antecedent rainfall average is 0.0045. Both numbers were the
same order of magnitude as those for TCP. The respective
probabilities for SWE are 0.018 and 0.022 (for 2.7-multiple of
the previous mean). Here the numbers are one order of magnitude larger than those for TCP. These results support the
Table 5
Weather conditions during Snow Thaw Period 2006 for selected stations
Weather station
LU
VI
ST
HL
VK
Elevation [m]
Snow Depth (SD) on March 20 [cm]
Snow Water Equivalent (SWE) on March 20/27[mm]
Discontinuous snow cover/Snow thawed completely [day]
Snow Thaw Rate (STR) from SWE [mm day 1]
Snow Thaw Rate (STR) from SD [cm day 1]
Snow Thaw Period (STP)
Rainfall in STP [mm]
Rainfall March 2629 [mm]
Mean daily temperature in STP [C]
Total Cumulative Precipitation (TCP) [mm]
Daily average TCP [mm day 1]
254
33
78.4/28.5
Mar 28/30
9.8
4.1
Mar 20Apr 4
65.1
50.6
NA
143.5
9.0
315
32
114.0/66.0
Mar 28/29
14.3
4.0
Mar 20Apr 4
49.1
41.9
5.0
163.1
10.2
315
40
137.0/55.0
Mar 29/30
15.2
4.4
Mar 22Apr 5
61.8
42.4
4.9
164.6
11.0
340
30
119.7/0
Mar 26/Apr 5
20.0
5.0
Mar 20Apr 2
57.3
50.0
NA
176.5
12.6
430
36
57.6/33. 2
Mar 30/31
5.8
3.6
Mar 20Apr 6
62.8
47.4
NA
120.4
7.5
252
Fig. 4. Relationship between landslide frequency (%) and the Total Cumulative
Precipitation (TCP; mm).
253