Document Actions

IJEGE-11_BS-Jiang-et-alii

background image
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
903
DOI: 10.4408/IJEGE.2011-03.B-098
DEBRIS FLOW AND LANDSLIDE FORECAST BASED ON GIS AND DOP-
PLER WEATHER RADAR IN LIANGSHAN PREFECTURE
Y.H. JIANG
(*)
, F.Q.WEI
(*)
, J.H. ZHANG
(**)
, B. DENG
(***)
& A.S. XU
(*)
(*)Institute of Mountain Hazards and Environment / Key Laboratory of Mountain Hazards and Surface Process, Chinese Aca-
demy of Sciences,Chengdu 610041, Sichuan, China
(**)Hainan Meteorological & Ecological Academy, Haikou 570203, Hainan, China
(***)Liangshan Meteorological Observatory, Xichang 615000, Sichuan, China
ficult to forecast debris flow and landslide accurately
because of their complicated formations mechanism.
The past forecasting researches mainly studied critical
precipitation of debris flow and landslide taking place.
Tan Wanpei found out the critical precipitation distri-
bution rules in 35 different debris flow ditches by clus-
ter analysis methods (t
an
et alii, 1989). According to
the relationship between the mean diurnal precipita-
tion intensity and its duration, Wilson established the
precipitation isolines and applied it in the debris flow
and the landslide forecast in San Francisco America
(w
ilson
, 1997). Corominas studied the relationship
between the precipitation and landslides in Liobregat
River basin and obtained the critical precipitation of
debris flow taking place in this region (C
oRominas
et
alii, 1999). Bell established the critical precipitation
coefficient of debris flow and landslide in Durban
region of South African (b
ell
et alii, 2000). Having
studied the relationship between the precipitation and
debris flows in Piedmont region of northeast Italy,
Pietro Aleotti determined the critical precipitation of
debris flows in this region (P
ietRo
et alii, 2004). All
of these methods forecast debris flow and landslide to
take place or not to by analyzing and erecting statisti-
cal relationship between the precipitation and debris
flow and landslide events. Since not sufficiently con-
sidering the function of underlying surface, it is very
difficult to determine the critical precipitation value of
debris flow and landslide in some large regions,.
This paper took Liangshan Yi Autonomous Pre-
ABSTRACT
Debris flow and landslide forecast is an important
means of disaster reduction. This paper took Liang-
shan Yi Autonomous Prefecture, Sichuan Province as
the study area. Analyzing distribution rules of debris
flow and landslide under different underlying sur-
face’s conditions, it used the extension theory to erect
to debris flow forecasting model and the information
content analysis method to erect the landslide forecast-
ing model. These models obtained 3 hours’ forecast-
ing precipitation by processing some Doppler weather
radar products. The debris flow and landslide forecast-
ing system in Liangshan Prefecture based on GIS was
put into use in Liangshan Meteorological Observatory
and could provide the future 3 hours’ debris flow and
landslide forecast. Experimental results showed that
the forecasting results have fine reliabilities.
K
ey
words
: debris Flow, landslide, forecast, GIS, Doppler
weather Radar
INTRODUCTION
Debris flow and landslide are the common disas-
ters in mountain regions. The two disasters have very
strong relativities. Almost all debris flows, introduced
by strong rain, take place with landslides (t
an
et alii,
1994; C. et alii, 1997). Debris flow and landslide intro-
duced by strong rain often burst in a sudden, so debris
flow and landslide forecast has become an important
means of disaster reduction. However, it is very dif-
background image
Y.H. JIANG, F.Q. wEI , J.H. ZHANG, B. DENG
& A.S. XU
904
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
sequence such as Xigeda strata. Yushu- Kangding-
Dongchuan earthquake belt influence the study area
greatly. The seismic intensity of Anning River Valley
in the middle of the study area is VIII~IX degree.
The study area belongs to the subtropics plateau
monsoon climate zone. The climate can be classified
as dry and wet seasons. The vertical climate differ-
ence is obvious. The mean diurnal precipitation is
about 1000 mm and the precipitation in rainy season
(May to October) accounts for 84-95%, especially, the
precipitation in the flood season (June to August) ac-
counting for about 60% of a year. The monthly mean
precipitation is more than 200 mm.
In a word, the geology, terrain and precipitation
in the study area provide pretty basic conditions for
debris flow and landslide growth. And then the study
area is one of the most serious debris flow and land-
slide disaster regions in China.
DEBRIS FLOw AND LANDSLIDE DISTRIBU-
TION IN THE STUDY AREA
There are 1105 debris flow valleys and 533 great
landslide points found in the study area. There are a
lot of debris flows (Fig. 1) and landslides (Fig. 2) in
most of counties. These two kinds of disasters mainly
distribute in the Intermediate Belt of big terrain, the
fault zone and the earthquake belt with rivers cutting
intensely, great relative height and rich precipitation.
These regions provide plenty of matter, energy and
rain for debris flow and landslide taking place.
Although, there are a little of debris flows in-
troduced by melting glacier water and landslides in-
fecture, Sichuan Province as the study area. Analyzing
the conditions of debris flow and landslide formation,
debris flow and landslide forecasting models were
erected respectively. The debris flow and landslide
forecasting system was developed on the GIS plat-
form in the study area. The system processed some
Doppler weather radar products to obtain importing
forecasting precipitation.
GENERAL SITUATION OF STUDY AREA
STATUS OF PHYSICAL GEOGRAPHY
Liangshan Yi Autonomous Prefecture situates
at the Southwest of Sichuan province. It is located
between longitudes 100°15' E and 103°53' E and
latitude 26°03' N and 29°27' N. It covers a total area
of 60,100 km
2
. The study area lies between the first
ladder and the second ladder of China’s terrain. The
highest elevation is 5958m and the lowest is only 305
m. Since the study area situates at the boundary be-
tween eastern platform region and western geosyn-
cline region and lies on the junction of the Pacific tec-
tonic domain and Tethyan tectonic domain, geologic
structures are complicated, new tectonic movements
are intense, and seismic activities is frequent. Many
issues of different direction's breaks are developed,
which demonstrates the geologic history is compli-
cated and the rock mass has been destructed by the
tectonic movements many times. The study area lo-
cates at the medium section of Kang-Dian old land
and many stratum are exposed. There are many kinds
of exposed rocks from the most ancient Presinian
such as metasandstone, slate, and phyllite to Tertiary
Fig. 1 - Distribution of debris flows in study area
Fig. 2 - Distribution of great landslides in study area
background image
DEBRIS FLOW AND LANDSLIDE FORECAST BASED ON GIS AND DOPPLER WEATHER RADAR IN LIANGSHAN PREFECTURE
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
905
three basic factors (material, energy, and rain) of de-
bris flow formation.
The loosing materials reserve is the best factor
expressing the physical conditions, but it is impos-
sible to obtain the accurate material reserve in each
forecasting unit. The loosing materials reserve is only
evaluated by its producing causes which mainly in-
cluding stratum-rock property (X1), fault density (X2)
and land use (X3). The energy condition includes rela-
tive height and slope which expresses the conversions
from potential energy to kinetic energy.
Since forecasting units have the same size, rela-
tive height is well correlated with slope in an unit. To
decrease calculation, the relative height (X4) is chosen
to express the energy factor. Precipitation (X5) and
precipitation intensity (X6) are chosen as evaluation
indexes of rain condition, the former provides suf-
ficient water source, and the latter provides dynamic
condition by forming formidable surface runoffs or
big interstitial hydraulic pressures.
In different cases of 6 predictors, there are differ-
ent probabilities B
j
(j=1, 2, … , 6) of debris flow tak-
ing place. In order to judge Bj under different combi-
natorial conditions R(X
1
, X
2
, …, X
6
) of predictors, the
standard matter-element model is set up firstly.
where x
1
, x
2
, …, x
6
are the values of X
1
, X
2
, …, X
6
.
Then the correlation degree, between any com-
binatorial state (R) of six factors and the probability
(Bj) of debris flow happening, can be calculated. The
correlation degrees of six factors under R and B
j
are
assumed as k(x).
where X
0
=< a b > X =< c d > X
0
X X
0
is the classical
region, and X is limited the region,
Assuming the correlation degree under R and Bj
troduced by other factors, most of debris flows and
landslides are related with high rainfall. Rain is the
primary factor introducing debris flows and land-
slides. Therefore, most of debris flows and landslides
in the study area take place in wet seasons (May to
October), especially in June to September.
DEBRIS FLOW AND LANDSLIDE FORE-
CASTING MODELS
THE FORECAST UNIT DIVISION
In order to analyze and operate forecast data con-
veniently, the study area was divided into many same
size units. The appropriate units are beneficial to the
accuracy of debris flow and landslide forecasting. The
oversized units could not express the real situation of
small terrain units or small valleys. Too small units
maybe only partly express circumstances of debris flow
valleys and landslide points. These two cases could re-
sult in wrong positives or wrong negatives to forecast.
The related research (w
ei
et alii, 1999) indicated
that the areas of more than 80% debris flow valleys
were within 10 km
2
, and the areas of approximate
85% active landslide points were within 10,000 m
2
.
Therefore, a debris flow forecasting unit was divided
into 3km×3km, and a landslide forecasting unit was
100m×100m.
THE EXPRESSION OF FORECAST RESULT
The current debris flow and the landslide forecast
is probability forecasting, however, restricted by the
complicating formation of debris flow and landslide
and the precision of basic data, it is impossible to ac-
curately determine definite probabilities forecasting.
So possibilities forecasting are expressed by differ-
ent probability intervals which often have five levels
(w
ei
et alii, 2004). From first level to fifth level, the
probability intervals are <0, 0.2>, <0.2, 0.4>, <0.4,
0.6>, <0.6, 0.8>, and <0.8, 1>. Among these levels,
from third level to fifth level express bigger prob-
abilities with yellow, orange, and red.
THE DEBRIS FLOw FORECASTING MODEL
The debris flow forecasting model was estab-
lished on basis of element theory and extension set
theory of Extenics (C
ai
, 1999; C
ai
et alii, 2000).
Fuzzy mathematics is used to transform qualitative
indexes to quantificational indexes. The evaluating
indicators in each unit would be chosen around the
(1)
(2)
background image
Y.H. JIANG, F.Q. wEI , J.H. ZHANG, B. DENG
& A.S. XU
906
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
spread. Therefore, the distance from a unit to rivers and
roads is chosen to express the free face condition.
In a forecasting unit, there are seven predictors
including stratum-rock property(A
1
), fault density(A
2
),
land use(A
3
),earthquake activity (A
4)
, relative height
(A
5
), the distance to rivers(A
6
) the distance to roads(A
7)
.
Assuming each predictor(A
ij
=1,2,...,7) has its differ-
ent state Aij (j=1,2,3,....r), and under different states of
these seven predictors, landslides taking place(B) are
possible. The importance of Aij to B can be expressed
by information quantity.
where
is the information quantity of B under state A
ij
, P(B/
A
ij
) is the probability of B under state Aij, and P(B) is
the general probability of B. When the value of
is greater, the predictor Ai under the state
j
can pro-
vide greater information quantity to B, and then land-
slides will take place more easily.
Assuming
where λ
i
is the weight of predictor A
i
. I is the total
information quantity. I is bigger, and then landslides
takes place more easily.
Because it is difficult to determine P(B/A
ij
), the
probability of landslides taking place while Ai under
state j, and P(B/A
i
), the general probability while Ai
under the combinatorial state, the formula (4) have
to be changed into the formula (6) according to the
probability multiplication theorem.
To simplify computation, while sample vol-
ume is enough, the probability can be replaced by
frequency. Then the formula (6) is changed into the
formula (7).
is Kj(P), and then
where α
i
is the weight of X
i
relative to other factors;
k
j
(x
i
) is the correlation degree under X
i
and B
j
.
According to the principle of most subjection de-
gree, the probability of debris flow happening can be
worked out by Formula (3). If k
k
(P) is equal to
the probability of debris flow taking place is Bk under
state R.
THE LANDSLIDE FORECAST MODEL
Landslides introduced by rain are the results of rain
acting on underlying surfaces with different conditions.
Then complex underlying surfaces may be looked as
the integrity. Firstly, analyze the sensitivities of land-
slides to underlying surfaces with different conditions.
Secondly, analyze the probability of landslide taking
place while different precipitation acting on different
underlying surfaces. At last determine the probability.
SENSITIVITY ANALYSIS OF LANDSLIDE IN FO-
RECASTING UNIT
The sensitivity of landslide was evaluated by the
information content analysis method. There are so
many environmental factors restricting and influencing
landslides happening, and different factors have differ-
ent effects. Essential conditions of material, energy and
free-face must be provided while landslides happening.
The material condition refers to the distributing con-
dition of material easily participating in the formation
of landslides. There is no effectual method to directly
obtain the distributing condition material in large re-
gions , so only those factors influencing broken degree
of rocks can be chosen to evaluate material conditions.
The primary factors include stratum-rock property,
fault density, earthquake activity, land use, and etc..
Land use mainly expresses the surface cover and the
influence of human activity. Landslides taking place
need some certain slope. Based on good relativity be-
tween slope and relative height in same size units, rela-
tive height was chosen to express the energy condition.
Both river erosions and side slopes excavation could
produce free faces. Among the formations of excava-
tion to side slope, road construction is the most wide-
(3)
(4)
(5)
(6)
(7)
background image
DEBRIS FLOW AND LANDSLIDE FORECAST BASED ON GIS AND DOPPLER WEATHER RADAR IN LIANGSHAN PREFECTURE
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
907
forecasting unit, so the size of final outputting should
be 3km ×3km.
A unit of 3km × 3km can be divided into 900 units
of 100m × 100m, and then to landslide forecast, the
value of every outputting unit adopts the greatest fore-
cast grade among its 900 units of 100m × 100m. The
greater grade between the debris flow result and the
landslide result is the final synthetic result in a unit.
PARAMETER DETERMINING IN FORE-
CAST MODELS
PARAMETER DETERMINING IN THE DEBRIS
FLOw FORECASTING MODEL
After extracting the information quantity of un-
derlying surface factor in every forecasting unit where
debris flow ditches locate, the distribution rule of de-
bris flow ditches in every underlying surface factor
can be analyzed. When the number of tests is bigger,
the frequency is closer to the probability. There are
7159 units in the study area, so the probability can be
replaced by the frequency like formula (10).
where p
i
is frequency of debris flow ditches having
the underlying surface factor i, N
i
is the number of de-
bris flow ditches having the underlying surface factor
i, and S
i
is the number of units having the underlying
surface factor i.
Fig. 3 is frequencies of debris flow ditches appear-
ing in the units with different relative heights. When
the relative height is below some value, the relative
height is greater and the frequency of debris flow tak-
ing place is greater. The reason is that the topographi-
cal condition is more propitious to debris flow taking
place with the relative height increasing. But when the
relative height is over some value, the tendency is re-
where N
ij
is the number of units where landslides
taking place while A
i
under state j, S
ij
is the number
of units with Ai having state j, N is the total number
of units landslides taking place, and N is the total
number of units in study area.
THE PROBABILITY OF LANDSLIDES TAkING
PLACE wHILE RAIN ACTING ON
Landslide is the result of rain acting on under-
lying surface. When the sensitivity of landslide to
underlying surface is determined, the probability of
landslide with different precipitation could be evalu-
ated. The total information quantity I of seven pre-
dictors and precipitation P are the final predictors of
landslide forecast.
The probability B
j
(j=1,2) of landslide is differ-
ent under different combination states of I and P. In
order to determine Bj under different combination
state R ( I,P) , a standard element model about the
probability of landslide can be constructed as for-
mula (8) firstly.
where m
1
is the value of I and m
2
is the value of P.
The same to computation of the correlation de-
gree in debris flow forecast model, the correlation
degree k
j
(R) of the probability B
j
under state R could
be computed as formula (9).
where αi is the weight of mi, k
j
(m
i
) is the correlation
degree of mi to Bj.
if
and then the probability of landslide under state R is
determined as B
k
.
SYNTHETICAL RESULT OF DEBRIS
FLOW AND LANDSLIDE FORECAST
Due to promulgating the disaster forecasting
grade to public not pointing out debris flow or land-
slide, forecasting results of debris flow and landslide
must be processed synthetically. The size of a debris
flow forecasting unit is bigger than that of a landslide
(8)
(9)
(10)
Fig.3 - frequencies of debris flow ditches in different rela-
tive heights
background image
Y.H. JIANG, F.Q. wEI , J.H. ZHANG, B. DENG
& A.S. XU
908
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
verse. With the slop increasing, it is difficult to reserve
loosing solid matter in slope-faces
Fig. 4 is frequencies of debris flow ditches appear-
ing in the units with different fault densities. The fault
density is greater, which makes loosing solid matter's
forming condition better, and the frequency of debris
flow taking place is greater.
Fig. 5 is frequencies of debris flow ditches appear-
ing in the units with different land use indexes (WEI et
alii
, 2008). The land use index is greater, which makes
the loosing solid matter's forming condition better, and
the frequency of debris flow taking place is greater.
Fig. 6 is frequencies of debris flow ditches appear-
ing in the units with different stratum-rock properties.
According to the hardness and weather-resistant abil-
ity, the stratum appearing in study area can be divided
into five grades. With the hardness and weather-resist-
ant ability decreasing, which makes loosing solid mat-
ter forming and accumulating easily, then debris flow
may take place easily.
According to the four trending figures, threshold
values of the four underlying surface factors can be
determined in every grade in the debris flow fore-
casting model.
PARAMETERS DETERMINING IN THE LAN-
DSLIDE FORECAST MODEL
The information quantity of every underlying sur-
face factor in landslide forecast can be computed by
formula (7). The total information quantity of under-
lying surface factors in every unit can be obtained by
formula (5).
Fig. 7 is frequencies of landslides appearing in the
units with different total information quantities. With
the information quantity increasing, the frequency of
landslide taking place is greater. According to this, the
threshold value of the total underlying surface factors
information quantities can be determined in every
grade in the landslide forecast model.
PRECIPITATION
OBTAINING
AND
ANALYZING
FORECASTING PRECIPITATION PRODUCTS
PROCESSING
Doppler weather radar could quickly catch
strong convective weather process and obtain high
precision inspecting and forecasting precipitation
products. The space resolution is 1 km. The radar
provides products dBZ (21#) and VIL (57#). From
these two products, forecasting precipitation and
one-hour rainfall intensity could be obtained. Meth-
ods are as follows.
Product 21# and 57# must be processed by geom-
etry transformation with formula (11) firstly.
Fig. 4 - frequencies of debris flow ditches in different fault
densities
Fig. 5 - frequencies of debris flow ditches in different land
use indexes
Fig. 6 - frequencies of debris flow ditches in different stra-
tum-rock properties
Fig. 7 - Frequencies of landslides in different total infor-
mation quantities
(11)
background image
DEBRIS FLOW AND LANDSLIDE FORECAST BASED ON GIS AND DOPPLER WEATHER RADAR IN LIANGSHAN PREFECTURE
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
909
The forecasting system calculated 3 hours’ fore-
casting precipitation (Fig. 8) and one-hour rainfall
intensity (Fig. 9) by Doppler weather radar inspect-
ing data at 22, July 14, 2006. And then the system
produced 3 hours’ disaster forecast (Fig. 10). The
where X and Y are ground coordinate, x and y are the
coordinates of the image, and A, B, C, D, E and F are
transforming parameters (k
anG
-
tsunG
, 2004).
Choose no less than three points (control points)
whose ground coordinates known in the Doppler weath-
er radar images. Put image coordinates and ground co-
ordinates of control points into formula (5), transform
equations into standard form, and then calculate least-
square solutions of six transforming parameters (A, B,
C, D, E and F). The six transforming parameters could
be used to set up mapping relationship between radar
image coordinates and ground coordinates
Among the forecasting precipitation and one-
hour rainfall intensity, dBZ (21#) accounts for 2/3,
and VIL (57#) accounts for 1/3. Then the two prod-
ucts plus weighted participate in calculation. The
forecasting precipitation and one-hour rainfall inten-
sity will be objective.
ANTECEDENT PRECIPITATION PROCESSING
Antecedent precipitation is a part of the rainfall
factors in debris flow forecast. It is provided by pre-
cipitation database. Antecedent precipitation must
be considered attenuation and then transformed into
effective antecedent rainfall (s
enoo
et alii, 2008) to
act in debris flow precipitation. The transforming
method adopts the follow formula (12) (1985).
where R
t
is the observational precipitation of t
antecedent day, a
t
= 0.5
t/T
is the attenuation rate of
rainfall and T is the half life of rainfall. Values of T
are different in different regions.
THE FORECASTING SYSTEM APPLYING CA-
SES
The debris flow and landslide forecasting system
was put into use in Liangshan Meteorological Ob-
servatory in Sichuan province in June, 2006.
CASE 1
There was an extremely big debris flow disas-
ter in Yanyuan County in the study area on July 14,
2006. The debris flow took place at about 23. This
was the only one debris flow aroused by rainstorm in
the region in 2006.
(12)
Fig.8 - 3 hours’ forecasting precipitation at 22 July 14,
2006
Fig. 9 - 3 hours’ forecasting one-hour rainfall intensity at
22 July 14, 2006
Fig
. 10 - Forecasting map of debris flow in study area at 22
July 14, 2006
background image
Y.H. JIANG, F.Q. wEI , J.H. ZHANG, B. DENG
& A.S. XU
910
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
a number of conclusions can be summarized as fol-
lows:
- The method of debris flow and landslide forecast-
ing, which was based on fuzzy mathematics, ex-
tension theory and information content theory by
analyzing debris flow and landslide forming fac-
tors and relationship of the two, could resolve the
problems of debris flow and landslide forecasting
perfectly.
- Owing to complications of debris flow and land-
slide forming, it was very difficult to forecast ex-
act probability of debris flow and landslide. So
it was more preferable and practical to use five
probability ranges to express debris flow and
landslide forecasting results.
- Doppler weather radar inspecting rain has its char-
acters such as data updating fast, high resolution
and continual space. These characters could pro-
vide powerful means of inspecting and forecast-
ing precipitation, and they could improve the ac-
curacy of debris flow and landslide forecasting.
- The forecasting system was put into use in study
area and obtained favorable effects. However the
density of precipitation observation stations was
very low, Doppler weather radar in study area
was set up not so long and accumulation data was
not long enough, so accuracy of debris flow and
landslide forecasting needed to be improved.
ACKNOWLEDGEMENTS
This research was financially supported by the
Research Fund for Commonweal Project (Meteorol-
ogy) (GYHY201006039).
result where debris flow took place was forth grade
(itsprobability was in the range from 0.6 to 0.8). This
indicated that the forecasting result was reliable.
CASE 2
There was a big debris flow disaster in Xichang
County in the study area on September 18, 2008. The
debris flow took place at 5 past 0.
The forecast system calculated 3 hours’ fore-
casting precipitation (Fig. 11) and one-hour rainfall
intensity (Fig. 12) by processing Doppler weather
radar inspecting products at 23, September 17,2008.
And then the system produced 3 hours’ disaster
forecast (Fig. 13). Most of the regions where debris
flow took place were forth grade and fifth grade (its
probability was in the range from 0.6 to 1).
CONCLUSIONS
Through studying the methods of debris flow
and landslide forecasting and analyzing the results,
Fig.11 - 3 hours’ forecasting precipitation at 23 Sep. 17,
2008
Fig.12 - 3 hours’ forecasting one-hour rainfall intensity at
23 Sep. 17, 2008
Fig. 13 - Forecast map of disasters in study area at 23 Sep.
17, 2008
background image
DEBRIS FLOW AND LANDSLIDE FORECAST BASED ON GIS AND DOPPLER WEATHER RADAR IN LIANGSHAN PREFECTURE
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
911
REFERENCES
a
leotti
P. (2004) - A warning system for rainfall-induced shallow failures. Engineering Geology, 73: 247-265.
b
ell
f.G. & m
aud
R.R. (2000) - Landslides associated with the colluvial soils overlying the Natal Group in the greater Durban
region of Natal, South Africa. Environmental Geology, 39(9): 1029-1038.
C
ai
w. (1999) - Extension theory and its application. Chinese Science Bulletin, 44(17): 1538-1548.
C
ai
w., y
anG
C.-y. & l
in
w.-C. (2000) - Extension Engineering Method. Beijing: Science Press.
C
HanG
k.-t. (2003) - Introduction to Geographic Information Systems. Beijing: Science Press.
C
oRominas
J. & m
oya
J. (1999) - Reconstructing recent landslide activity in relation to rainfall in the Liobregat River basin,
Eastern Pyrenees, Spain. Geomorphology, 30(1-2): 79-93.
l
ee
C.f. & C
Hen
H. (1997) - Landslide in Hong kong-Causes and Prevention. ACTA Geographic Sinica, 52(5): 114~122.
s
enoo
k., G
odai
H. & H
aRa
y. (2000) - Rainfall indexes for debris flow warning evacuating program. Shin-Sabo 38(2): 16-21.
t
an
w. (1985) - Distribution characters of critical rainfall line for debris flow gully. Bulletin of Soil and Water Conservation,
1989, 9(6): 21-26.
t
an
w. & w
anG
C. (1994) - Zonal Forecast of Rainfall induced Debris Flow and Landslide. Sichuan Sciences and Technique
Press.
w
ei
f. & x
ie
H. (1999) - The fuzzy information model of debris flow risk factor. Journal of Chinese Soil and Water Conservation.
30(4): 273-277.
w
ei
f., t
anG
J. & x
ie
H. (2004) - Debris Flow Forecast Combined Regions and Valleys and Its Application. Journal of Mountain
Science, 22(3): 321-325.
w
ei
f., k
eCHanG
G. & k
aiHenG
H. (2008) - Relationships between debris flows and earth surface factors in Southwest China.
Environmental Geology.
w
ilson
R.C. (1997) - Normalizing rainfall / debris-flow thresholds along the U. S. Pacific coast for long-term variations in pre-
cipitation climate. i
n
C
Hen
C.L (ed.), Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment, Proceedings
1
st
International Conference on Debris-Flow Hazard Mitigation, ASCE, New York: 32-43.
Statistics