Professional Documents
Culture Documents
Debris Flow PDF
Debris Flow PDF
Engineering Geology
Rainfall thresholds for the occurrence of debris flows in the Jiangjia Gully,
Yunnan Province, China
Jianqi Zhuang a,b,⁎, Peng Cui c, Gonghui Wang d, Xiaoqing Chen c, Javed Iqbal e, Xiaojun Guo c
a
School of Geological Engineering and Surveying of Chang'an University, Key Laboratory of Western China Mineral Resources and Geological Engineering, Xi'an 710054, China
b
Institute of Geo-hazards Mitigation of Chang'an University, Xi'an 710054, China
c
Institute of Mountain hazards and Environment, CAS, Chengdu 610064, China
d
Research Centre on Landslides, Disaster Prevention Research Institute, Kyoto University, Gokasho, Uji, Kyoto 611-0011, Japan
e
Department of Earth Sciences, COMSATS Institute of Information Technology, Abbottabad, Pakistan
a r t i c l e i n f o a b s t r a c t
Article history: The Jiangjia Gully (JJG), Yunnan Province, China, is one of several debris flow gullies that characterize the topog-
Received 24 September 2014 raphy of mountainous areas in China. Rainfall-triggered debris flows are frequent in the JJG. In order to model and
Received in revised form 10 June 2015 predict these debris flow occurrences, we analyzed precipitation parameters (including intensity–duration
Accepted 14 June 2015
thresholds and rainfall event–duration thresholds) that triggered past debris flow events in the JJG. Since the
Available online 17 June 2015
JJG is unique, any model used for predicting debris flows in this gully must be developed using data from only
Keywords:
this gully. Furthermore, the effects of normalized precipitation and antecedent precipitation on debris flows in
Debris flow the JJG were considered. Debris flow data and precipitation data were compiled from 47 debris flow events in
Prediction 1993, 1994, 1998–2001, 2004 and 2006 in JJG. All of these debris flows were triggered following a precipitation
Rainfall thresholds event with a duration of 6 h or less, with most of the precipitation events exhibiting a duration of 3 to 6 h. Only
Jiangjia Gully three rainfall events with a duration of less than 1 h and average intensity between 1.0 and 42.0 mm/h produced
debris flows. The 90%, 70% and 50% probability curves for debris flow events were then constructed. Intensity–
duration (I–D) and rainfall event–duration (RE–D) thresholds were then used for debris flow modeling based
on preliminary tests and comparisons. Antecedent precipitation was found not to be a significant factor in trig-
gering debris flows in the JJG; however, intraday precipitation played a significant role. Normalized precipitation
threshold curves from adjacent areas were not well-correlated with the patterns observed in the JJG. Determina-
tion of unique thresholds for each gully is necessary for developing an effective prediction system.
Crown Copyright © 2015 Published by Elsevier B.V. All rights reserved.
1. Introduction Guzzetti et al., 2007, 2008; Baum and Godt, 2010; Engel et al., 2011;
Jakob et al., 2012). Many researchers have attempted to correlate
Debris flows occur in various mountainous regions around the world, rainfall/precipitation with debris flows (Wieczorek and Guzzetti, 1999;
with effects on local economies, as well as on life and infrastructure (e.g. Guzzetti et al., 2008). Accordingly, many critical thresholds and models
dams, roads) in nearby counties and villages. They are characterized by for debris flow initiation have been suggested using different precipita-
mass densities ranging from 1.3 to 2.3 t/m3, speeds of 3 to 15 m/s, flow tion parameters, such as precipitation duration–precipitation intensity
discharges from tens to thousands of times the normal stream flow in (I–D model) (Cannon et al., 2008, 2011), antecedent rainfall–cumulative
the region, and a wide range in particle diameters, from 10−6 m to rainfall (AR–R model) (Wieczorek and Guzzetti, 1999), antecedent
10 m (Cui et al., 2005; Zhuang et al., 2010). Rainfall, especially that from effective rainfall–cumulative rainfall (AER–R model) (Baum and Godt,
short duration rainstorms, has been identified to be one of the most 2010), maximum hourly rainfall intensity–cumulative rainfall (Imax–R
important factors responsible for triggering debris flows (Wieczorek model) (Guzzetti et al., 2007, 2008), normalized rainfall–cumulative rain-
and Guzzetti, 1999; Cui et al., 2005; Chen et al., 2006a,2006b; Guzzetti fall (Inorm–R model) (Guzzetti et al., 2008), and daily rainfall–maximum
et al., 2007, 2008; Zhuang et al., 2009; Shieh et al., 2009; Floris et al., hourly rainfall intensity (Imax–Rd model) (Guzzetti et al., 2007, 2008).
2010; Cannon et al., 2011; Li et al., 2011; Hu et al., 2011; Guo et al., The effective critical threshold for a precipitation-induced landslide
2013). Therefore, it is vital to forecast the occurrence of such flows and must take into account the local topography, geology, debris characteris-
develop effective warning systems based on the relationships between tics (distribution, volume and soil strength) and hydrogeological parame-
precipitation and debris flow events (Crosta, 1998; Aleotti, 2004; ters (Cannon, 1988; Guzzetti et al., 2007, 2008; Cannon et al., 2008; Baum
and Godt, 2010; Cannon et al., 2011). Among the critical threshold models
⁎ Corresponding author at: Institute of Geo-hazards Mitigation of Chang'an University, that have been proposed by researchers around the world, the corre-
Xi'an 710054, China. sponding rainfall durations/intensities differ by up to three orders of
http://dx.doi.org/10.1016/j.enggeo.2015.06.006
0013-7952/Crown Copyright © 2015 Published by Elsevier B.V. All rights reserved.
336 J. Zhuang et al. / Engineering Geology 195 (2015) 335–346
magnitude between certain models (Guzzetti et al., 2007, 2008). Even lack the necessary resources for the provision of continuous support
within the same geographical zone, there can exist differences in rainfall or expansion of services to other areas due to variation in local geolog-
duration and/or intensity of up to one order of magnitude; this variation ical conditions.
arises from rainfall sampling resolution, and physiographical, geological Rainfall-induced debris flows are relatively common in the Jiangjia
or geomorphological differences. In other words, a limitation of regional Gully (JJG), Yunnan Province, China. Located in mountainous south-
or even local precipitation thresholds is the fact that thresholds defined western China, the debris flows in this gully are fairly typical of those
for a particular area cannot be easily implemented for assessing surround- that occur in other mountainous areas in China. Regular observations
ing areas, including some areas that possess similar environmental condi- and systematic data collection of debris flows in JJG have been conduct-
tions (Crosta and Frattini, 2008; Crosta, 1998). ed by the Institute of Mountain Hazards and Environment at the Chinese
The most common precipitation parameters used in determining Academy of Sciences since 1965. Typically, 10 to 20 min of high-
critical thresholds for debris flows are antecedent rainfall, antecedent intensity rainfall can initiate a debris flow from shallow landslides in
effective rainfall, precipitation duration, precipitation intensity, cumula- this gully, with an average debris flow density of 2.0 t/m3 (Kang et al.,
tive rainfall, maximum hourly rainfall intensity, normalized rainfall, and 2004; Cui et al., 2005). This is due to the presence of complex geological
daily rainfall (Guzzetti et al., 2008). Rainfall thresholds can be grouped structures, fragile rock formations, numerous shallow landslides, and
into two broad categories; (i) empirical threshold, which uses historical abundant rainfall in the gully. During every rainy season (May–Octo-
debris flow and precipitation data (Guzzetti et al., 2007, 2008; Baum ber), about 12–20 debris flows occur in the region (Fig. 1). Several
and Godt, 2010; Zhuang et al., 2014); (ii) physical threshold, which prediction models have been proposed with respect to the JJG based
employs a hydro-mechanical model of debris flow occurrences on intraday rainfall (Wei et al., 2005; Cui et al., 2008) and antecedent
(Takahashi, 1991; Cui, 1992; Iverson, 1997). Several decades of research 10-min rainfall (Wu et al., 1990; Guo et al., 2013) before debris flow
on debris flow initiation processes suggests that the former method is occurrences, but none of them have demonstrated acceptable accuracy.
more extensively used worldwide (Guzzetti et al., 2007; Baum and Meanwhile it is still unclear what the most influential rainfall parame-
Godt, 2010). Empirical rainfall thresholds are defined by measuring ters (antecedent rainfall, intraday rainfall and rainfall intensity) are, in
rainfall events that have resulted in debris flows; the most commonly terms of the triggering of debris flows. In this study, we developed
used units of measurement for the parameters of these rainfall events prediction models using different rainfall parameters and methods,
include: (i) Intraday rainfall and antecedent effective rainfall (Tan and and then validated the models using actual debris flow events. Based
Duan, 1995; Wei et al., 2005; Cui et al., 2008; Zhuang et al., 2014), (ii) on this, the influences of different rainfall parameters on debris flow
rainfall intensity–duration thresholds (Godt et al., 2006; Guzzetti occurrence are discussed.
et al., 2007; Cannon et al., 2008), (iii) cumulative rainfall event–duration In order to forecast debris flows in the JJG, the characteristics of past
thresholds (Kanji et al., 2003; Aleotti, 2004; Giannecchini, 2005), (iv) precipitation events that triggered debris flows in the gully were
cumulative rainfall event–average rainfall intensity thresholds analyzed to construct a comprehensive precipitation threshold model,
(Heyerdahl et al., 2003; Hong et al., 2005), and (v) other thresholds, including both I–D and RE–D models. Data was obtained from
including normalized cumulative rainfall events and normalized rainfall DCDFORS (Dongchuan Debris Flow Observation and Research Station,
intensity (Cannon, 1988; Jibson, 1989; De Vita et al., 1998; Jakob and located in Yunnan Province, China; Fig. 2). Finally, the effects of different
Weatherly, 2003). Numerous studies have analyzed the empirical precipitation parameters on debris flow occurrences are discussed.
relationship between rainfall and debris flow occurrences, proposing
various critical thresholds and equations for different regions. For exam- 2. Geological setting of the study area
ple, studies were conducted near San Francisco, CA, USA (Baum and
Godt, 2010), wildfire-prone areas in the western USA (Cannon et al., The Jiangjia Gully is a tributary of the Xiaojiang River, which has 106
2008), the Alps in Europe (Guzzetti et al., 2007), Taiwan (Chen et al., other tributaries that are also subject to debris flows. The watershed of
2006a, 2006b; Shieh et al., 2009; Kung et al., 2012), Hong Kong (Chan the Xiaojiang River is colloquially known in China as the “World's
et al., 2003) and the Xiaojiang–Anning river basin in China (Tan and Natural Museum of Debris Flows” (Cui et al., 2005). The drainage
Duan, 1995; Wei et al., 2005; Cui et al., 2008; Hu et al., 2010). However, basin intersects the Xiaojiang River at a distance of 31 km south of the
the agencies responsible for implementing safety protocols generally junction of the Xiaojiang River and the Jinsha River. The JJG has a
30 700
Number of debris flow
Sediment transport 600
25
Nember of debris flow occurrence
500
20
400
15
300
10
200
5
100
0 0
1965 1970 1975 1980 1985 1990 1995 2000 2005
Year
Fig. 1. The number of debris flow events and annual sediment transport by debris flows during an average rainy season in the JJG, Yunnan, China.
J. Zhuang et al. / Engineering Geology 195 (2015) 335–346 337
Yinjiaao Precipitation
Observing Station
Soil infiltration
test site Mayiping Precipitation
Observing Station
Beijing
Second-Yangtze
terrace River
Third-
Kunming terrace
Fig. 2. Map of the study area and the observation site (a: location of the JJG in China; b: observation area of the JJG; c: debris flow event locations; d: watershed of the JJG and observation
stations).
trunk channel length of 13.9 km and covers a total area of 48.6 km2. It is their color (light and dark, respectively). Both rock types are weak,
located within the Xiaojiang fault zone in northeastern Yunnan Province and easily weathered and fragmented. Colluvium is widely distributed
in Southwest China (N23°13′–23°17′, E103°6′–103°13′). on slopes, which provide materials for shallow landslides. The total
The study area is characterized by intense tectonism; about 80% of cumulative volume of clastic detritus, which acts as potential source
the exposed rocks are highly fractured and slightly metamorphosed. material for future debris flows, is estimated to be 1.2 × 109 m3 (Wu
The predominant sandstone and slate can be easily differentiated by et al., 1990). The main stream extends from the drainage basin at an
Precipitation
Precipitation
altitude of 3269 m, westwards to the junction with the Xiaojiang River The annual average sediment yield in the JJG is 2.0 million m3, while
at an altitude of 1042 m. Slopes steeper than 25° comprise 55% of the a maximum of 6.6 million m3 was reported in 1991 as a result of
basin area, with an altitude difference of 500 m (Wu et al., 1990) (Fig. 2). unusually numerous and heavy debris flows in that year (Cui et al.,
During every rainy season (May–October), when 85% of the annual 2005). The average annual rainfall is 800 mm, with a maximum of
precipitation falls, about 12–20 debris flows occur in the region. The 1130 mm and a minimum of 514 mm (during the period 1953–2010).
record for the highest annual number of events was 28 in 1965. In Thunderstorms are frequent during the rainy season. Typically, a
1961, DCDFORS was established with the cooperation of the Institute high-intensity rainfall event lasting 10 to 20 min can trigger a debris
of Mountain Hazards and Environment at the Chinese Academy of flow (Kang et al., 2004). The discharge rate of each debris flow ranges
Sciences. Since 1987, regular observations and systematic data collec- from hundreds to thousands of cubic meters per second (Kang et al.,
tion on debris flows have been conducted. 2004).
Table 1
Debris flow and corresponding precipitation data (debris flow peak discharge (Pd), debris flow discharge (Di), sediment transport (St), debris flow density (De)).
100.0
Curve 90%
Curve 70%
Curve 50%
10.0 D
Intensity (mm/h)
1.0
A C
B
0.1
0.1 1.0 10.0 100.0
Duration (h)
Fig. 4. The relationship between rainfall duration/intensity and the 90%, 70% and 50% probability thresholds for the initiation of a debris flow event.
3. Data collection and manipulation such as intensity (I), duration (D), antecedent effect rainfall (AER),
and normalized intensity (I-normal).
3.1. Precipitation data The definition of a rainfall event has been discussed in other studies.
For example, Jan and Lee (2004) defined it to be no more than 4 mm of
Three rain-gauges were installed in JJG in 1993, one each in the rainfall within 1 h before the rainfall event, while Zhou and Tang (2013)
Menqian Gully and Duozhao Gully, and one at DCDFORS. These rain- defined it as no more than 1 mm within 6 h before the rainfall event. As
gauges provided high-resolution rainfall data (with a resolution of discussed earlier, debris flows in the JJG are frequent and can be trig-
0.1 mm in precipitation and 1 min in observation intervals). In 2002, gered by low-intensity rainfall events. Accordingly, we defined a rainfall
an additional six rain-gauges were installed; three at the same sites as event here as the precipitation measured from the beginning of a storm
those previously indicated, and three in the Menqian Gully. Previous to the occurrence of a debris flow. The beginning of a storm is defined as
studies (Wu et al., 1990; Cui et al., 2005; Hu et al., 2010; Guo et al., the occurrence of no more than 0.1 mm rainfall during 3 h (or more)
2013) have indicated that precipitation in the Menqian Gully is the prior to a storm. The rainfall event duration is defined as the time that
main trigger for debris flows in the JJG, due to the presence of unconsol- elapses from the beginning of a storm to the occurrence of a debris
idated, loose materials and debris flow initiation form shallow flow. The intensity is defined as the total rainfall in the event divided
landslides. Accordingly, data from the Mayiping Rain Gauge was used by the duration (Fig. 3).
as the main data source. Antecedent effective rainfall is defined here as the actual effect of
For a given setting, debris flows are typically triggered by a defined rainfall on soil water content prior to a rainfall event that triggered a
set of rainfall conditions. These conditions may involve high rainfall debris flow. However, the role of antecedent rainfall in triggering debris
intensities over short durations and/or lower rainfall intensities flows diminishes to zero as the interval between the two increases, due
maintained over longer durations. Therefore, it is necessary to analyze to soil water redistribution (infiltration) and evaporation processes
the relationships between debris flow and various rainfall parameters, (Bruce and Clark, 1969; Ni et al., 2010; Li et al., 2011). Thus, the amount
100
D C
Curve 50%
Rainfall event (mm)
Curve 90%
Curve 70% B
10
1
0 1 10 100
Duration (h)
Fig. 5. The relationship between rainfall duration/event and the 90%, 70% and 50% probability thresholds, calculated using Eq. (3).
340 J. Zhuang et al. / Engineering Geology 195 (2015) 335–346
1.00
More frequently
Intensity (mm/h)
0.10
0.01
1 10 100
Rainfall event (mm)
of water infiltrating into hillslopes during a rainfall preceding a It should be noted that a normalized rainfall event is defined as the
landslide event can be referred to as the antecedent effective rainfall rainfall divided by mean annual precipitation; normalized duration is
for the debris flow event (Glade et al., 2000). Therefore, it is appropriate defined as the rainfall event duration divided by mean annual precipita-
to calculate this parameter, as it considers the actual contribution of tion; normalized mean intensity is defined as the mean intensity
antecedent rainfall to debris flow occurrences. The antecedent effective divided by mean annual precipitation (Cannon, 1988; Jibson, 1989; De
rainfall is commonly considered a function of the regressive rainfall Vita et al., 1998; Jakob and Weatherly, 2003). Finally, rainfall intensity
index and time (Bruce and Clark, 1969): is defined as the rainfall duration (h) divided by the rainfall amount
(mm) (Cannon, 1988).
X
n
Re ¼ Ri K i ð1Þ
i¼1
3.2. Debris flow data
where Re represents the antecedent effective rainfall (mm) that fell
during i days before the debris flow (or similar geological event) Since 1993, more than 70 debris flows have been recorded in the JJG.
occurred; K is the degressive index representing the outflow of the However, the debris flows occurring in 2002, 2003, and 2005 were ex-
regolith; Ri is the intraday rainfall before the debris flow occurred. cluded from the dataset due to issues with rain-gauge data collection
Eq. (1) calculates the antecedent effective rainfall via analysis of statistics, during these years. Hence, only 60 debris flow events and correspond-
and indicates the effective precipitation before a rainfall-triggered debris ing precipitation data were included for analysis. The data include the
flow. We used the soil moisture content attenuation rate to determine date of each debris flow, debris flow peak discharge (Pd), debris flow
the number of days preceding the debris flow occurrence. The data discharge (Di), sediment transport (St), debris flow density (De),
used in this study confirm that the influence of rainfall diminishes and precipitation (with a resolution of 0.1 mm), rainfall duration (with a
eventually disappears with an increasing time interval (due to evapora- resolution of 1 min), intensity, and effective antecedent rainfall before
tion), and has no appreciable influence on debris flow formation if 15 each debris flow event (Table 1). We analyzed 47 debris flow occur-
or more days have elapsed. Hence, the upper limit of n in Eq. (1) is 15. rences in 1993, 1994, 1998–2001, 2004 and 2006 to determine our
The degressive index (K) has been previously studied by many re- thresholds. We then used 13 debris flows in the period 2007–2008 to
searchers; we used a value of 0.78 (Cui et al., 2008; Zhuang et al., 2009). validate these thresholds.
100
More frequently
Rainfall event (mm)
10
1
1 10 100
Antecedent effective rainfall (mm)
Fig. 7. The relationship between rainfall events and effective antecedent precipitation.
J. Zhuang et al. / Engineering Geology 195 (2015) 335–346 341
1.00
More frequently
Intensity (mm/h)
0.10
0.01
1 10 100
Antecedent effective rainfall (mm)
Fig. 8. The relationship between rainfall intensity and effective antecedent precipitation.
4. Results Using Eq. (2), we calculated the best-fit line using the least squares
and medium-scale methods, and obtained the equation I = 15.87
4.1. I–D threshold D−0.5952. This equation uses median regression to empirically obtain
the corresponding median thresholds (curve 50% of the graph for
Thresholds empirically derived from rainfall intensity–duration including 50% of the debris flow event points). It only demonstrates
have been widely used to identify rainfall conditions that result in the the relationship between rainfall intensity and duration, and cannot
occurrence of debris flows (Guzzetti et al., 2007, 2008; Baum and be used directly to forecast debris flows using only one probability
Godt, 2010; Lee et al., 2013). Inspection of the I–D thresholds reveals line (“yes” or “no” with respect to debris flow occurrences). Accordingly,
the general form: we calculated the 90% and 70% (curve 90%/70% of the graph for
including 90%/70% of the debris flow event points) probability curves
in the graph, corresponding to 90% and 70% of the debris flow event
I ¼ c þ αDβ ð2Þ
points, respectively (i.e. the lowest threshold line and the lowest medi-
an threshold line, respectively) (Fig. 4). The equations to calculate the
where I is (mean) rainfall intensity, D is rainfall duration, and c, α and β 70% and 90% thresholds are I = 11.92 D−0.5952 and I = 7.27 D−0.5952,
are other parameters. For the majority of I–D thresholds, c = 0 and respectively. The quadrant can be divided by the three thresholds into
Eq. (2) takes the form of a simple power law.To model the empirical four zones (A, B, C and D). Zone A is located below the lowest threshold
relationship between I and D, the two variables were plotted on a single (90% curve) and represents a very low probability of debris flow. Zone B
graph (Fig. 4), where D (x − axis) is the rainfall duration and I (y-axis) is located between the lowest threshold line (90% curve) and lowest
is the rainfall intensity. About 70% of the storms that generated debris median threshold line (70% curve), representing a relatively low
flows in the JJG had a duration equal to or less than 6 h, with most probability of a debris flow occurrence for points located between the
durations being 3 to 6 h. I and D curves. This would necessitate the release of a watch alert.
100.0 7 3
5
8
Intensity (mm/h)
10.0
4 1
10 2 9 6
50% Curve
70% Curve
This study Test data 1 90% Curve
1.0
2 3 4
5 6 7
8 9 90% Curve
Fig. 9. Comparison and validation of I–D thresholds (No. 1: Caine (1980); No. 2: Wieczorek (1987); Nos. 3, 4, 5: Jibson (1989); No. 6: Guadagno (1991); No. 7: Paronuzzi et al. (1998); No. 8:
Crosta and Frattini (2001); No. 9: Shieh et al. (2009); No. 10: Guo et al. (2013)).
342 J. Zhuang et al. / Engineering Geology 195 (2015) 335–346
1000
4
2
1
0.1 1.0 10.0 100.0
Duration (h)
Fig. 10. Comparison and validation of RE–D thresholds (No. 1: Innes (1983); Nos. 2–3: Wilson et al. (1992); No. 4: Kanji et al. (2003)).
Zone C is located between the lowest median threshold line (70% curve) RE b 16 mm), using the 20% debris flow occurrence (P = 20%) curve.
and the median threshold line (50% curve), representing a fairly high This line marked the boundary between the frequent and infrequent
probability of debris flow occurrence. This would require the issuance categories.
of a warning alert. Zone D is located above the upper median threshold
line, indicating an extremely high probability of debris flow occurrence RE ¼ 17:45 D0:3042 for curve 50%
(Fig. 4). RE ¼ 13:98 D0:3042 for curve 70% ð3Þ
RE ¼ 8:12 D0:3042 for curve 90%
4.2. RE–D and RE–I thresholds
Besides the I–D threshold, thresholds based on events described by 4.3. I–AER and RE–AER thresholds
total rainfall and rainfall duration (Campbell, 1975), named RE–D
thresholds (Cannon and Ellen, 1985), have also been utilized to predict Soil moisture is a factor that predisposes slopes to failure and then
rainfall-induced debris flow. Caine (1980) also suggested certain the use debris flow occurrences (Crozier, 1986; Wieczorek, 1996). In order to
of global RE–I thresholds for debris flows using accumulate rainfall understand the effect of soil moisture on debris flow occurrences, we
event and rainfall intensity. The rainfall event required to trigger devised a simple method based on antecedent effective precipitation
landslides in the JJG varied from a minimum of 3.5 mm to a maximum to establish a threshold.
of 75 mm. As indicated in Fig. 5, the RE–D thresholds for different debris Relationships involving I–AER (Fig. 7) and RE–AER (Fig. 8) were
flow occurrence probabilities is defined by the relationship between RE analyzed. The AER needed to trigger landslides in JJG varied from a
and D. However, it is difficult to quantify the relationship between RE minimum of 12 mm to a maximum of 35 mm. Exploring more complex
and I (Fig. 6). The rainfall frequencies for different debris flow occur- relationships between the AER and the RE (I) threshold curves are
rence probabilities were classified in our study into frequent (I N 0.08 beyond the scope of this paper. The demarcation between frequent
mm/min and RE N 16 mm) and infrequent (I b 0.08 mm/min and and infrequent occurrences of debris flows can be defined (20%
0.10
2
3 1
6
I-Normal (mm)
8
5
0.01
Fig. 11. The I-normal-D thresholds in the JJG and their comparison with other threshold models (Nos. 1–6: Jibson (1989); No. 7: Paronuzzi et al. (1998); No. 8: Wieczorek et al. (2000)).
J. Zhuang et al. / Engineering Geology 195 (2015) 335–346 343
Debris flow no-occurs therefore, the focus of these models is on minimizing these anomalies
(Guzzetti et al., 2008; Zhuang et al., 2014). In the future, this model
High Debris flow occurs High
Threshold line can be revised and calibrated by including data on subsequent debris
flows and precipitation.
Precipitation
Loose debris
proposed in different studies. In addition to significant variation in
their parameters, all except curves 4, 17 and 20 are higher than the
one we developed. Thresholds 4 and 17 were constructed based on
Precipitation Loose debris areas affected by wildfires, while threshold 20 was developed based
on the Chi–Chi earthquake (Shieh et al., 2009). As with the JJG, both
Low
Low areas are prone to debris flows at low precipitation thresholds, because
Time of the presence of abundant loose debris.
Fig. 9 indicates the various thresholds proposed for the JJG in
High Debris flow occurs High different studies. The I–D 90% threshold proposed by Guo et al. (2013)
Threshold line
(No. 21) is lower than the 90% threshold developed by us. The reason
Precipitation
Loess debris
for this discrepancy is that the definitions of rainfall events that
triggered debris flows are different. Guo et al. (2013) defined a sudden
Debris flow no-occurs change in rainfall as critical rainfall, and defined the rainfall duration
threshold as 1 h and 10 min. This was used as the basis for modeling
Precipitation in their study.
Fig. 10 indicates the four RE–D threshold curves that have been
Low Low proposed in different studies. Threshold curve 1, which is the lowest of
Time the four, is comparable to ours. Curve 2, developed in a study on volcanic
debris flows in Hawaii, is slightly higher than the 90% threshold curve
Fig. 12. Debris flow initiation characteristics (top: debris-limited; bottom: rainfall-limit- presented in our paper. The remaining two threshold curves are signifi-
ed). Revised from Bovis and Jakob (1999).
cantly higher than the ones presented here. Reviewing Figs. 9–10, it can
be concluded that the thresholds for debris flow occurrences vary signif-
probability), for AER–RE, as AER = 13.5 mm and RE = 19.5 mm, while icantly across local, regional and global scales, due to differences in geo-
for AER–I, it can be defined as AER = 12 mm and RE = 0.05 mm/min. logical, morphological and climatic conditions. Therefore, it is necessary
to develop a unique model for each gully subject to major debris flows.
5. Discussion
5.2. Normalization threshold
5.1. Model testing and comparison
As discussed, a limitation of local I–D thresholds is the fact that those
In order to verify the reliability and accuracy of the determined defined for a particular area with its unique climatic and geological
thresholds and make necessary refinements, we used data from 13 conditions cannot be easily implemented for surrounding areas
debris flow events in the period 2007–2008 for model validation. (Figs. 9 and 10). This is mainly due to differences in the rainfall param-
Fig. 9 indicates the model validation results using the I–D thresholds. eters that trigger debris flows. To compare rainfall thresholds prepared
Eleven debris flows were located in the ‘B’ region and two were located for different regions, several authors used the concept of normal or nor-
in the ‘C’ region. Fig. 10 indicates the model validation results using the malized rainfall intensity, which is the rainfall intensity divided by the
RE–D thresholds. Ten and two debris flows were located in the‘B’ and ‘C’ mean annual precipitation (Cannon, 1988; Jibson, 1989; De Vita et al.,
regions, respectively, while one was located in the ‘D’ region. The results 1998; Jakob and Weatherly, 2003). In view of ongoing climate change,
indicate that the RE–D threshold line is slightly more accurate than the we used a similar factor, albeit refined further by using annual precipi-
I–D threshold line in the JJG. The occurrence of false negatives and false tation within the same year as the divisor. The generated relationship
positives is not unprecedented in even highly accurate models; is referred to as I-normal-D.
40 30
rainfall 10cm
20cm 30cm
25
40cm
30
Soil moisture (%)
20
Rainfall (mm)
20 15
10
10
5
0 0
6-23 7-1 7-6 7-10 7-19 7-25 8-5 8-9 8-13 8-21 8-25
Date
Fig. 13. Soil moisture at different soil depths, plotted against precipitation in the JJG in 2006.
344 J. Zhuang et al. / Engineering Geology 195 (2015) 335–346
12
4
20cm
0
0 5 10 15 20 25 30
Time (min)
In order to verify whether the I-normal-D threshold was applicable that characterize the two types of flows are different. This is yet another
to other regions, we analyzed their relationship in the JJG and compared reason why the rainfall thresholds that trigger debris flows are unique
it with previously presented models (Fig. 11). We observed the follow- to one or a few regions.
ing: (i) The I-normal-D curves vary significantly across studies; (ii) the
I-normal-D curves obtained in our study are lower than those presented 5.3. The effect of AER on debris flows in the JJG
by other researchers; (iii) use of normalized thresholds cannot avoid
variability arising from geographical, morphological, geological, and Current literature on rainfall thresholds for predicting landslide
certain climatic factors, or variability tied to the amount of source hazards suggests that there are two viewpoints regarding the selection
material distributed near and in the flow channels. Hence, the challenge of appropriate precipitation parameters; one focuses on the rainfall in-
in extrapolating these thresholds to other locations remains. tensity (Guzzetti et al., 2007) and the other on the antecedent effective
The debris flows that typically occur in the JJG can be classified as rainfall (Cui et al., 2008). The influence of rainfall on landslides differs
rainfall-limited flows. Debris flows initiated by landslides due to heavy substantially depending upon the landslide type and materials involved
precipitation are different from those caused by bed failure due to (Guzzetti et al., 2008). Shallow failures are usually triggered by short,
runoff, with different underlying mechanisms (Guzzetti et al., 2007, intense storms, while deep-seated failures are triggered by long-term
2008; Gregoretti and Fontana, 2008; Zhuang et al., 2013). Debris flows antecedent effective rainfall (Li et al., 2008; Piacentini et al., 2012;
can be categorized as either rainfall-limited (mainly initiated by land- Chen et al., 2014). Debris flows can be initiated by either shallow or
slides) or debris-limited (mainly caused by bed failure due to runoff), deep-seated landslides. A study on tropical soils of Hong Kong revealed
with different trigger factors. This is the case even when geographical, that antecedent rainfall of any duration is not significant in the trigger-
morphological and climatic conditions are similar (Bovis and Jakob, ing process (Brand et al., 1984). On the other hand, a similar study in
1999). The former is characterized by a practically unlimited sediment Wellington, New Zealand, revealed that rainfall during antecedent
supply and is triggered by a critical value of precipitation. However, in periods of up to 10 days influenced the soil moisture balance with
the case of debris-limited flows, a substantial interval must elapse for respect to triggering deep landslides (Crozier, 1999). This effect of ante-
channel recharge and sediment supply to build up before the next cedent moisture may be related to regional climates, soil permeability
debris flow can occur. The initiation of such flows is affected by both properties and the debris flow initiation type (shallow landslide, deep
precipitation and debris supply (Fig. 12). In addition, the loose materials landslide or run-off).
0.7 35
0.6 30
0.5 25
Rainfall/ mm
Runoff /m3/s
0.4 20
Base flow in JJG
0.3 15
0.2 10
0.1 5
0 0
8/1/09 8/5/09 8/9/09 8/13/09 8/17/09 8/21/09 8/25/09 8/29/09
Date
Fig. 15. Runoff in the JJG after a rainfall event in Aug. 2009.
J. Zhuang et al. / Engineering Geology 195 (2015) 335–346 345
In the JJG, antecedent precipitation has been considered to be a (3) Although precipitation was normalized according to climatic
significant factor in the triggering of debris flows, according to other parameters, the normalized precipitation threshold curve cannot
researchers; the length of the effective antecedent precipitation period be directly implemented to surrounding areas with similar
is 15 to 20 days (Cui et al., 2005, 2008; Wu et al., 1990; Guo et al., 2013). conditions, due to differences in topography.
In this paper, we analyzed the I-AER (Fig. 7) and RE-AER (Fig. 8) (4) Debris flows in JJG are predominantly triggered by intraday
relationships, and found that the two were not well-correlated. Fig. 13 precipitation, while antecedent precipitation is not a significant
indicates the soil moisture at different soil depths, plotted against factor.
precipitation in JJG in 2006. Soil moisture was measured by TDR
(Time-Domain Reflectometry) in a grassland near DCDFORS (see the Acknowledgments
Fig. 2). Soil moisture was highly sensitive and responsive to precipita-
tion, with minimal lag effects. In addition, the water contents at We are thankful to Prof. Hong Yong and Hu Kaiheng at the Institute
different depths varied in their responses to precipitation. Although of Mountain Hazards and Environment, CAS for their valuable sugges-
the water contents at depths of 10 cm and 20 cm were responsive, tions and assistance with this study. The authors sincerely appreciate
those at depths of 30 and 40 cm were not. In addition, soil moisture the assistance with data collection provided by Dr. Lin Yongming, Dr.
was strongly correlated to rainfall on the same day, with decreasing Ge Yonggang and Dr. Zhu Xinghua from the Institute of Mountain
correlation coefficients as the interval after the preceding rainfall Hazards and Environment, CAS and Chen Wenbo from the CEE of the
event lengthened. When the interval reached four days, soil moisture PolyU of HongKong. We express special thanks to the staff of Dongchuan
tended to decrease to its original equilibrium value (Fig. 13). In other Debris Flow Observation and Research Station of the Institute of Moun-
words, elevated soil water content was present for only 4 days after a tain Hazards and Environment, CAS, for their cooperation. This work
rainfall event. Therefore, antecedent precipitation did not significantly was financially supported by the National Natural Science Foundation
affect the soil water content. of China (41202244) and the National Key Fundamental Research
Fig. 14 shows the water infiltration rate in a landslide body, measured Program of China (973) (2014CB744703).
using a DRI (Double Ring Infiltrometer), within the Menqian Gully
(Fig. 2). The infiltration rate was observed to be high initially (up to 15 References
cm/min), and then decreased sharply to achieve a steady infiltration
Aleotti, P., 2004. A warning system for rainfall-induced shallow failures. Eng. Geol. 73,
rate of 0.4 cm/min. This value is higher than that observed in other
247–265. http://dx.doi.org/10.1016/j.enggeo.2004.01.007.
areas, probably because the JJG comprises loose, unconsolidated Baum, R.L., Godt, J.W., 2010. Early warning of rainfall-induced shallow landslides and
materials and witnesses high evapotranspiration rates (3752.7 mm/y) debris flows in the USA. Landslides 7, 259–272.
Bovis, M.J., Jakob, M., 1999. The role of debris flow conditions in predicting debris flow
(Cui et al., 2005). Runoff in JJG in Aug. 2009 was measured by DCDFORS
activity. Earth Surf. Process. Landf. 24, 1039–1054.
using CSM (Cross-Section Measurement) (Fig. 15). That indicates that Brand, E.W., Premchitt, J., Phillipson, H.B., 1984. Relationship between rainfall and
the runoff response to precipitation was evident and rapid, increasing landslides in Hong Kong. Proc. 4th Int. Symp. on Landslides, vol. 1. Toronto,
sharply with precipitation and decreasing sharply after precipitation pp. 377–384.
Bruce, J., Clark, R., 1969. Introduction to hydrometeorology. Pergamon Press, London,
terminated. Soil water content tended to decrease to the initial value pp. 252–270.
within days of the precipitation event. Caine, N., 1980. The rainfall intensity-duration control of shallow landslides and debris
The high infiltration rate indicates that the soil has little water-holding flows. Geogr. Ann.Ser. A: Phys. Geogr. 62, 23–27.
Campbell, R.H., 1975. Soil slips, debris flows, and rainstorms in the Santa Monica
capacity. This is due to insufficient vegetation cover and the prevalence of Mountains and vicinity, southern California. US Geological Survey Professional
coarse materials in the slope of the gully Infiltrating water converges as Paper 851. U.S. Government Printing Office, Washington DC, p. 51.
underground water flow and flow away via macropores. Meanwhile, Cannon, S.H., 1988. Regional rainfall-threshold conditions for abundant debris-flow
activity. Landslides, floods, and marine effects of the storm of January 3–5, 1982, in
debris flows in the JJG are primarily initiated by shallow landslides (Wu the San Francisco Bay region, California. U.S. Geological Survey Professional Paper
et al., 1990; Cui et al., 2005; Chen et al., 2006a, 2006b; Hu et al., 2011). (No. 1434).
These landslides can be triggered when an extended period of heavy Cannon, S.H., Ellen, S.D., 1985. Rainfall conditions for abundant debris avalanches, San
Francisco Bay region, California. Calif. Geol. 38, 267–272.
precipitation occurs, concurrent with a maximum depth of water
Cannon, S.H., Gartner, J.E., Wilson, R.C., Bowers, J.C., Labe, J.L., 2008. Storm rainfall
infiltration of 2 m in the slope body. Under such conditions, a shallow conditions for floods and debris flows from recently burned areas in southwestern
landslide may progress into a debris flow. Therefore, the soil water Colorado and southern California. Geomorphology 96, 250–269.
Cannon, S.H., Boldt, E.M., Laber, J.L., Kean, J.W., Staley, D.M., 2011. Rainfall intensity–
and run-off in the JJG is usually not independent of antecedent
duration thresholds for postfire debris-flow emergency-response planning. Nat.
precipitation. Hazards 59 (1), 209–236.
Chan, R.K.S., Pang, P.L.R., Pun, W.K., 2003. Recent developments in the landslip warning
6. Conclusion system in Hong Kong. Proceedings of the 14th Southeast Asian Geotechnical
Conference. Balkema, Lisse, The Netherlands, pp. 7–84.
Chen, H., Dadson, S., Chi, Y.G., 2006a. Recent rainfall-induced landslides and debris flow in
In this study, we analyzed the rainfall conditions that triggered northern Taiwan. Geomorphology 77 (1), 112–125.
debris flows in JJG in the past, in order to develop rainfall intensity– Chen, X.Q., Cui, P., Feng, Z.L., Chen, J., Li, Y., 2006b. Artificial rainfall experimental study on
landslide translation to debris flow. Chin. J. Rock Mech. Eng. 25 (1), 106–116.
duration thresholds and rainfall event–duration thresholds determining Chen, S.C., Chou, H.T., Chen, S.C., Wu, C.H., Lin, B.S., 2014. Characteristics of rainfall-
the occurrence of debris flows in the JJG. These two thresholds were val- induced landslides in Miocene formations: a case study of the Shenmu Watershed,
idated using 2007–2008 debris flows and then compared with other Central Taiwan. Eng. Geol. 164, 133–146.
Crosta, G., 1998. Regionalization of rainfall thresholds: an aid to landslide hazard
thresholds determined from around the world. Finally, the effects of evaluation. Environ. Geol. 35 (2–3), 131–145.
normalized precipitation and antecedent precipitation on debris flows Crosta, G.B., Frattini, P., 2001. Rainfall thresholds for triggering soil slips and debris flow.
were discussed, with the following conclusions: In: Mugnai, A., Guzzetti, F., Roth, G. (Eds.), Proc. 2nd EGS Plinius Conf. on Mediterra-
nean Storms, pp. 463–487 (Siena).
Crosta, G.B., Frattini, P., 2008. Rainfall-induced landslides and debris flows. Hydrol.
(1) The I–D and RE–D thresholds for the 50%, 70% and 90% probabil- Process. 22 (4), 473–477.
ities of debris flows were determined. Subsequently, four regions Crozier, M.J., 1986. Landslides: causes, consequences and environment. Croom Helm,
London, p. 252.
corresponding to different warning levels were identified. These
Crozier, M.J., 1999. Prediction of rainfall-triggered landslides: a test of the antecedent
thresholds are applicable in predicting future debris flows in the water status model. Earth Surf. Process. Landf. 24, 825–833.
JJG. Cui, P., 1992. Studies on condition and mechanism of debris flow initiation by means of
(2) The threshold curves proposed in other areas indicated in this experiment. Chin. Sci. Bull. 37, 759–763.
Cui, P., Chen, X., Wang, Y.Y., Hu, K., Li, Y., 2005. Jiangia Ravine debris flows in
manuscript tend to be higher than the ones determined here, south-western China. In: Jakob, M., Hungr, O. (Eds.), Debris-flow hazards and related
and are therefore not applicable to the JJG. Phenomena. Springer, Heidelberg, pp. 565–594.
346 J. Zhuang et al. / Engineering Geology 195 (2015) 335–346
Cui, P., Zhu, Y.Y., Chen, J., 2008. Relationships between antecedent rainfall and debris Natural Slopes and Earthfills. Iw-Flows 2003 2. Associacione Geotecnica Italiana,
flows in Jiangjia Ravine, China. In: Chen, C.L., Rickenmann, D. (Eds.), Proceedings of Sorrento, pp. 1–11.
the Fourth International Conference on Debris Flow. Springer Press, Wellington, Kung, H.Y., Chen, C.H., Ku, H.H., 2012. Designing intelligent disaster prediction models and
pp. 1–9. systems for debris-flow disasters in Taiwan. Expert Syst. Appl. 39 (5), 5838–5856.
De Vita, P., Reichenbach, P., Bathurst, J.C., Borga, M., Crosta, G., Crozier, M., Wasowski, J., Lee, D.H., Lai, M.H., Wu, J.H., Chi, Y.Y., Ko, W.T., Lee, B.L., 2013. Slope management criteria
1998. Rainfall-triggered landslides: a reference list. Environ. Geol. 35 (2), 219–233. for Alishan Highway based on database of heavy rainfall-induced slope failures. Eng.
Engel, Z., Česák, Julius, Escobar, V.R., 2011. Rainfall-related debris flows in Carhuacocha Geol. 162, 97–107.
Valley, Cordillera Huayhuash, Peru. Landslides 8, 269–278. Li, C.J., Ma, S.H., Zhu, X.S., 2008. Forecasting of landslide triggered by rainfall: thoery,
Floris, M., D'Alpaos, A., Squarzoni, C., Genevois, R., Marani, M., 2010. Recent changes in methods and applications. Geological Publishing House 3–18.
rainfall characteristics and their influence on thresholds for debris flow triggering Li, C.J., Ma, T.H., Zhu, X.S., Li, W., 2011. The power–law relationship between landslide
in the Dolomitic area of Cortina d'Ampezzo, north-eastern Italian Alps. Nat. Hazards occurrence and rainfall level. Geomorphology 130 (3–4), 221–229. http://dx.doi.
Earth Syst. Sci. 10 (3), 571–580. org/10.1016/j.geomorph.2011.03.018.
Giannecchini, R., 2005. Rainfall triggering soil slips in the southern Apuane Alps (Tuscany, Ni, H.Y., Zheng, W.M., Li, Z.L., 2010. Recent catastrophic debris flows in Luding county, SW
Italy). Adv. Geosci. 2, 21–24 (SRef-ID: 1680-7359/adgeo/2005-2-21). China: geological hazards, rainfall analysis and dynamic characteristics. Nat. Hazards
Glade, T., Crozier, M.J., Smith, P., 2000. Applying probability determination to refine 55, 523–542. http://dx.doi.org/10.1007/s11069-010-9545-2.
landslide-triggering rainfall thresholds using an empirical. Antecedent Daily Rainfall Paronuzzi, P., Coccolo, A., Garlatti, G., 1998, Eventi meteorici critical debris flows nei bacini
Model. Pure Appl. Geophys. 157 (6/8), 1059–1079. http://dx.doi.org/10.1007/ montani del Friuli. L'Acqua, Sezione I=Memorie, pp, 39–50.
s000240050017. Piacentini, D., Troiani, F., Soldati, M., Notarnicola, C., Savelli, D., Schneiderbauer, S., Strada,
Godt, J.W., Baum, R.L., Chleborad, A.F., 2006. Rainfall characteristics for shallow landslid- C., 2012. Statistical analysis for assessing shallow-landslide susceptibility in South
ing in Seattle, Washington, USA. Earth Surf. Process. Landf. 31, 97–110. http://dx. Tyrol (south-eastern Alps, Italy). Geomorphology 151–152, 196–206.
doi.org/10.1002/esp.1237. Shieh, C.L., Chen, Y.S., Tsai, Y.J., Wu, J.H., 2009. Variability in rainfall threshold for debris
Gregoretti, C., Fontana, G.D., 2008. The triggering of debris flow due to channel-bed failure flow after the Chi-Chi earthquake in central Taiwan, China. Int. J. Sediment Res. 24
in some alpine headwater basins of the Dolomites: analyses of critical runoff. Hydrol. (2), 177–188.
Process. 22 (13), 2248–2263. Takahashi, T., 1991. Debris flow. Balkema, pp. 22–56.
Guadagno, F.M., 1991. Debris flows in the Campanian volcaniclastic soil (Southern Italy). Tan, B.Y., Duan, A.Y., 1995. Study on prediction for rainstorm debris flow along mountain
Proc. Int. Conf. on slope stability. Thomas Telford, Isle of Wight, pp. 125–130. district railways. J. Nat. Disasters 4 (2), 43–52.
Guo, X.J., Cui, P., Li, Y., 2013. Debris flow warning threshold based on antecedent rainfall: a Wei, F.Q., Hu, K.H., Chen, J., 2005. Determination of effective antecedent rainfall for debris
case study in Jiangjia Ravine, Yunnan, China. J. Mt. Sci. 10 (2), 305–314. flow forecast. J. Mt. Sci. 23 (4), 453–457.
Guzzetti, F., Peruccacci, S., Rossi, M., Stark, C.P., 2007. Rainfall thresholds for the initiation Wieczorek, G.F., 1987. Effect of rainfall intensity and duration on debris flows in central
of landslides in central and southern Europe. Meteorog. Atmos. Phys. 98, 239–267. Santa Cruz Mountains. In: Costa, J.E., Wieczorek, G.F. (Eds.), Debris flow and
Guzzetti, F., Peruccacc, S., Ross, M., Stark, C.P., 2008. The rainfall intensity–duration control avalanches: process, recognition, and mitigationReviews in Engineering Geology 7.
of shallow landslides and debris flows: an update. Landslides 5, 3–17. Geological Society of America, pp. 93–104.
Heyerdahl, H., Harbitz, C.B., Domaas, U., Sandersen, F., Tronstad, K., Nowacki, F., Engen, A., Wieczorek, G.F., 1996. Landslide Triggering Mechanisms. Landslides Investigation and
Kjekstad, O., Devoli, G., Buezo, S.G., Diaz, M.R., Hernandez, W., 2003. Rainfall induced Mitigation. In: Turner, A.K., Schuster, R.L. (Eds.), Transportation Research Board Na-
lahars in volcanic debris in Nicaragua and El Salvador: practical mitigation. Proc. Int. tional Academcy of Sciences, Special Report 247, pp. 76–90.
Conf. on Fast Slope Movements-Prediction and Prevention for risk Mitigation, Wieczorek, G.F., Guzzetti, F., 1999. A review of rainfall thresholds for triggering landslides.
IC-FSM2003. Patron Pub, Naples, pp. 275–282. Proc. of the EGS Plinius Conference, Maratea, Italy, pp. 407–414.
Hong, Y., Hiura, H., Shino, K., Sassa, K., Suemine, A., Fukuoka, H., Wang, G., 2005. The Wieczorek, G.F., Morgan, B.A., Campbell, R.H., 2000. Debris flow hazards in the Blue Ridge
influence of intense rainfall on the activity of large-scale crystalline schist landslides of Central Virginia. Environ. Eng. Geosci. 6, 3–23.
in Shikoku Island, Japan. Landslides 2 (2), 97–105. http://dx.doi.org/10.1007/s10346- Wilson, R.C., Torikai, J.D., Ellen, S.D., 1992. Development of rainfall thresholds for debris
004-0043-z. flows in the Honolulu District, Oahu. U.S. Geol. Surv. Open File Rep. 92–521, 45.
Hu, K.H., Cui, P., Wang, C.C., Li, Y., Lu, X.B., 2010. Characteristic rainfall for warning of de- Wu, J., Kang, Z., Tian, L., Zhang, S., 1990. Observation and research of debris flow in
bris flows. J. Mt. Sci. 7 (3), 207–214. http://dx.doi.org/10.1007/s11629-010-2022-2. Jiangjiagou Ravine, Yunnan Province. Science Pressing, Beijing, pp. 67–145.
Hu, M., Wang, R., Shen, J., 2011. Rainfall, landslide and debris flow intergrowth relation- Zhou, W., Tang, C., 2013. Rainfall thresholds for debris flow occurrence in the Wenchuan
ship in Jiangjia Ravine. J. Mt. Sci. 8 (4), 603–610. earthquake area. Adv. Water Sci. 24 (6), 786–793.
Innes, J.L., 1983. Debris flows. Prog. Phys. Geogr. 7, 469–501. Zhuang, J.Q., Cui, P., Ge, Y.G., Hong, Y., 2009. Relationship between rainfall characteristics
Iverson, R.M., 1997. The physics of debris flows. Rev. Geophys. 35 (3), 245–296. and total amount of debris flow. J. Beijing For. Univ. 31 (4), 77–83.
Jakob, M., Weatherly, H., 2003. A hydroclimatic threshold for landslide initiation on the Zhuang, J.Q., Cui, P., Ge, Y.G., He, Y.P., Liu, Y.H., Guo, X.J., 2010. Probability assessment de-
North Shore Mountains of Vancouver, British Columbia. Geomorphology 54 (3), bris flow block the main river associate Wenchuan Earthquake. Int. J. Remote Sens. 31
137–156. (13), 3465–3478.
Jakob, M., Owen, T., Simpson, T., 2012. A regional real-time debris-flow warning system Zhuang, J.Q., Cui, P., Peng, J.B., Hu, K.H., Iqbal, J., 2013. Initiation process of debris flows on
for the District of North Vancouver, Canada. Landslides 9 (2), 165–178. different slopes due to surface flow and trigger-specific strategies for mitigating
Jan, C.D., Lee, M.H., 2004. A debris flow rainfall-based warning model. J. Chin. Soil Water post-earthquake in old Beichuan County, China. Environ. Earth Sci. 68 (5),
Conserv. 35 (3), 275–285. 1391–1403.
Jibson, R.W., 1989. Debris flow in southern Porto Rico. Geol. Soc. Am. Spec. Pap. 236, Zhuang, J.Q., Iqbal, J., Peng, J.B., Liu, T.M., 2014. Probability prediction model for landslide
29–55. occurrences in Xi'an, Shaanxi Province, China. J. Mt. Sci. 11 (2), 345–359.
Kang, Z.C., Li, Z.F., Ma, A., Luo, J.T., 2004. Research on debris flow in China. Science Press,
pp. 12–66.
Kanji, M.A., Massad, F., Cruz, P.T., 2003. Debris flows in areas of residual soils: occurrence
and characteristics. Int. Workshop on Occurrence and Mechanisms of Flows in