# IJEGE-11_BS-Jiang-et-alii

*Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza*

*DOI: 10.4408/IJEGE.2011-03.B-098*

**DEBRIS FLOW AND LANDSLIDE FORECAST BASED ON GIS AND DOP-**

**PLER WEATHER RADAR IN LIANGSHAN PREFECTURE**

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

*et alii*, 1989). According to

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

River basin and obtained the critical precipitation of

debris flow taking place in this region (C

*et*

*alii*, 1999). Bell established the critical precipitation

coefficient of debris flow and landslide in Durban

region of South African (b

*et alii*, 2000). Having

debris flows in Piedmont region of northeast Italy,

Pietro Aleotti determined the critical precipitation of

debris flows in this region (P

*et alii*, 2004). All

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,.

**ABSTRACT**

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**

strong relativities. Almost all debris flows, introduced

by strong rain, take place with landslides (t

*et alii,*

*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-

*Y.H. JIANG, F.Q. wEI , J.H. ZHANG, B. DENG*

*& A.S. XU*

*5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011*

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.

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.

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

TION IN THE STUDY AREA

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.

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*

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

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*

**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*

bris flow formation.

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.

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.

*B*

*j*

*(j=1, 2, … , 6)*of debris flow tak-

natorial conditions R

*(X*

*1*

*, X*

*2*

*, …, X*

*6*

*)*of predictors, the

*x*

*1*

*, x*

*2*

*, …, x*

*X*

*1*

*, X*

*2*

*, …, X*

*6*

*.*

*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*

*k(x)*.

*X*

*0*

*=< a b > X =< c d >*

*X*

*0*

*X X*

*X*is limited the region,

*R*and

*Bj*

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*

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.

*et alii*, 1999) indicated

were within 10 km

into 3km×3km, and a landslide forecasting unit was

100m×100m.

*THE EXPRESSION OF FORECAST RESULT*

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

*et alii*, 2004). From first level to fifth level, the

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*

theory of Extenics (C

*et alii*, 2000).

indexes to quantificational indexes. The evaluating

indicators in each unit would be chosen around the

*Y.H. JIANG, F.Q. wEI , J.H. ZHANG, B. DENG*

*& A.S. XU*

*5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011*

roads is chosen to express the free face condition.

*A*

*1*

*A*

*2*

*A*

*3*

*A*

*4*)

*A*

*5*

*A*

*6*

*A*

*7*)

*A*

*ij*

*=1,2,...,7*) has its differ-

*Aij*(j=1,2,3,....r), and under different states of

these seven predictors, landslides taking place(B) are

possible. The importance of A

*ij*to B can be expressed

by information quantity.

is the information quantity of B under state

*A*

*ij*

*,*

*P*(

*B/*

*A*

*ij*

*B*under state

*Aij,*and

*P*(

*B*) is

*B*, and then land-

slides will take place more easily.

*A*

*i*

*.*I is the total

*I*is bigger, and then landslides

takes place more easily.

*P*(

*B*/

*A*

*ij*

state j, and

*P*(

*B*/

*A*

*i*

to be changed into the formula (6) according to the

probability multiplication theorem.

frequency. Then the formula (6) is changed into the

formula (7).

*j*(P), and then

*α*

*i*

*X*

*i*

*relative to other factors;*

*k*

*j*

*(x*

*i*

*)*is the correlation degree under

*X*

*i*

*B*

*j*

*.*

worked out by Formula (3). If

*k*

*k*

*(P)*is equal to

state R.

*THE LANDSLIDE FORECAST MODEL*

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

RECASTING UNIT

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-

**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*

be 3km ×3km.

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

FLOw FORECASTING MODEL

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).

*p*

*i*

*i*,

*N*

*i*

*is the number of de-*

*i*, and

*S*

*i*

*i.*

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-

*N*

*ij*

*A*

*i*

*under state*

*j*,

*S*

*ij*

*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

PLACE wHILE RAIN ACTING ON

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.

*B*

*j=1,2*) of landslide is differ-

*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.

*m*

*1*

*I*and

*m*

*2*

*P.*

degree

*k*

*j*

*R*) of the probability

*B*

*j*

*R*could

*αi*is the weight of mi,

*k*

*m*

*i*

*Bj*.

*R*is

determined as

*B*

*k*

**SYNTHETICAL RESULT OF DEBRIS**

**FLOW AND LANDSLIDE FORECAST**

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

*Fig.3 - frequencies of debris flow ditches in different rela-*

*tive heights*

*Y.H. JIANG, F.Q. wEI , J.H. ZHANG, B. DENG*

*& A.S. XU*

*5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011*

loosing solid matter in slope-faces

density is greater, which makes loosing solid matter's

forming condition better, and the frequency of debris

flow taking place is greater.

*et*

alii, 2008). The land use index is greater, which makes

alii

the loosing solid matter's forming condition better, and

the frequency of debris flow taking place is greater.

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.

determined in every grade in the debris flow fore-

casting model.

*PARAMETERS DETERMINING IN THE LAN-*

DSLIDE FORECAST MODEL

DSLIDE FORECAST MODEL

lying surface factors in every unit can be obtained by

formula (5).

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

PROCESSING

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.

*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*

**DEBRIS FLOW AND LANDSLIDE FORECAST BASED ON GIS AND DOPPLER WEATHER RADAR IN LIANGSHAN PREFECTURE**

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

coordinates of the image, and A, B, C, D, E and F are

transforming parameters (k

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

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*

cipitation database. Antecedent precipitation must

be considered attenuation and then transformed into

effective antecedent rainfall (s

*et alii*, 2008) to

method adopts the follow formula (12) (1985).

*R*

*t*

*t*

*a*

*t*

*T*is the half life of rainfall. Values of

*T*

are different in different regions.

*THE FORECASTING SYSTEM APPLYING CA-*

SES

SES

servatory in Sichuan province in June, 2006.

*CASE 1*

2006. The debris flow took place at about 23. This

was the only one debris flow aroused by rainstorm in

the region in 2006.

*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*

*Y.H. JIANG, F.Q. wEI , J.H. ZHANG, B. DENG*

*& A.S. XU*

lows:

- The method of debris flow and landslide forecast-

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.

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.

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.

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**

ogy) (GYHY201006039).

(itsprobability was in the range from 0.6 to 0.8). This

indicated that the forecasting result was reliable.

*CASE 2*

debris flow took place at 5 past 0.

intensity (Fig. 12) by processing Doppler weather

radar inspecting products at 23, September 17,2008.

flow took place were forth grade and fifth grade (its

probability was in the range from 0.6 to 1).

**CONCLUSIONS**

*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*

**DEBRIS FLOW AND LANDSLIDE FORECAST BASED ON GIS AND DOPPLER WEATHER RADAR IN LIANGSHAN PREFECTURE**

**REFERENCES**

*A warning system for rainfall-induced shallow failures*. Engineering Geology,

**73:**247-265.

*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.

*Extension theory and its application.*Chinese Science Bulletin,

**44**(17)

**:**1538-1548.

*Extension Engineering Method. Beijing*: Science Press.

*Introduction to Geographic Information Systems.*Beijing: Science Press.

*Reconstructing recent landslide activity in relation to rainfall in the Liobregat River basin,*

*Eastern Pyrenees,*

*Spain*. Geomorphology,

**30**(1-2)

**:**79-93.

*Landslide in Hong kong-Causes and Prevention.*ACTA Geographic Sinica,

**52**(5)

**:**114~122.

*Rainfall indexes for debris flow warning evacuating program*. Shin-Sabo

**38**(2)

**:**16-21.

*Distribution characters of critical rainfall line for debris flow gully.*Bulletin of Soil and Water Conservation,

**9**(6)

**:**21-26.

*Zonal Forecast of Rainfall induced Debris Flow and Landslide.*Sichuan Sciences and Technique

*The fuzzy information model of debris flow risk factor.*Journal of Chinese Soil and Water Conservation.

**30**(4)

**:**273-277.

*Debris Flow Forecast Combined Regions and Valleys and Its Application*. Journal of Mountain

**22**(3)

**:**321-325.

*Relationships between debris flows and earth surface factors in Southwest China.*

**(1997) -**

*Normalizing rainfall / debris-flow thresholds along the U. S. Pacific coast for long-term variations in pre-*

*cipitation climate.*i

*Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment*, Proceedings