CN103399344B - After a kind of earthquake there is the Forecasting Methodology of position in Rockfall hazard - Google Patents

Abstract

The Forecasting Methodology of Rockfall hazard position is there is after the invention provides a kind of earthquake, the method includes, and obtains multiple fundamental geological parameters in space to be predicted, multiple geologic parameters, seismologic parameter, the geographical position of at least one known slump point and the geographical position of multiple non-slump point；Newmark displacement D is calculated based on Newmark displacement model_n；With predicted position from river distance, from fracture belt distance, topographic relief amplitude and Newmark displacement D_nBuild binary Logistic regression equation；By described known slump point and each parameter of non-slump point and calculated Newmark displacement D_nAs known conditions, calculate the partial regression coefficient of each Variable Factors in regression equation；Calculated partial regression coefficient is utilized to build the Rockfall hazard forecast model in this space to be predicted.

Description

After a kind of earthquake there is the Forecasting Methodology of position in Rockfall hazard

Technical field

The present invention relates to the risk assessment technology field of disaster reduction and prevention and disaster after earthquake, more particularly it relates to an there is the Forecasting Methodology of Rockfall hazard position after earthquake.

Background technology

The safety of life and property of the mankind, as a kind of destructive disaster, is all caused great threat by earthquake.Violent earthquake, while above ground structure is damaged, also usually brings out secondary disaster.Particularly in mountain area, earthquake usually triggers a large amount of Secondary Geological Hazards, and the disaster such as these avalanches, landslide, mud-rock flow has increased the weight of earthquake loss, the direct losses that its loss caused causes sometimes even over earthquake itself.Avalanche after shake, landslide disaster are the class geological disasters triggered at first after the earthquake that continues, and have also established basis, thing source for mud-rock flow, barrier lake disaster simultaneously.Avalanche after shake, landslide disaster both can be produce with the generation of earthquake simultaneously, it is also possible to be just occur after certain time after earthquake.

China is positioned at circum-Pacific seismic belt and the intersection of Mediterranean-Himalaya earthquake zone, is one of multiple state of world's earthquake.Meanwhile, China is also the country on mountain more than, and mountain region, hills area account for the 70% of area.Many shakes superpose with many mountains so that the Secondary Geological Hazards such as the avalanche of China's earthquake-induced, landslide take place frequently, and lose huge, harm seriously.From between 1500-1949, precise record is had to produce secondary avalanche, the earthquake of landslide disaster has 134 times.China is one of country of avalanche in the world, landslide, mud-rock flow especially severe, and the direct economic loss caused because of geological disaster every year accounts for more than the 20% of natural disaster total losses.It turns out that, occur the violent earthquake in Southwestern China mountain area often to cause large-scale avalanche or landslide, cause serious economic loss and casualties.On May 12nd, 2008, Wenchuan violent earthquake has caused the Secondary Geological Hazards such as ten hundreds of avalanche, landslide, cause painful loss to disaster area people's life and property safety, and within the quite a long time, people's lives and properties and communal facility are yet suffered from safely grave danger.According to estimates, the geological disaster point that Wenchuan earthquake triggers has 3～40,000 places, based on avalanche, landslide, Rolling Stone.The investigation of Yin Yue equality people finds, Wenchuan earthquake induces nearly 15000 places avalanche, landslide, mud-stone flow disaster, result in about 20000 people dead.Sometimes, the Loss of Geological Hazards of Earthquake triggering is also bigger than the loss that earthquake itself causes, for instance, on JIUYUE 7th, 2012, the victim's great majority in the earthquake of 5.7 grades, the Yiliang of Yunnan Province Shao Tongshi are the rockfall hazards dying from Earthquake triggering.Occur the Lushan earthquake in Ya'an, Sichuan Province also to trigger substantial amounts of avalanche, landslide disaster on April 20th, 2013, block and cut off traffic, having had a strong impact on the transport of rescue personnel and goods and materials.Visible, the Rockfall hazard after earthquake is the Secondary Geological Hazards that Southwest Mountainous Areas can not be ignored.

Landslide and avalanche are usually accompanied, they have the contact that cannot split, result from identical geological tectonic environment with under identical formation lithology structural environment, and have identical triggering factors, the area easily producing landslide is also the Yi Faqu of avalanche, and therefore landslide type avalanche or avalanche type landslide are also referred to as slump stream.Avalanche, landslide can bring out mutually under certain condition, convert mutually.Avalanche and landslide that in context, earthquake causes also are called for short Rockfall hazard.

Understand the distribution characteristics of Rockfall hazard after shaking post-disaster reconstruction and risk assessment is all significant.At present, have a lot for the research of Rockfall hazard distribution characteristics after shake, main method is by field investigation and remote sensing image interpretation, obtain landslide distribution information after shaking, and study the landslide distribution situation under varying environment background based on GIS, statistical analysis is the distribution situation on landslide under the condition such as different intensity area, different elevation, different gradient, different lithology, explores its regularity of distribution.After shake, the research of Rockfall hazard distribution characteristics is that further exploration discovery earthquake-landslide influence factor has established important foundation.Existing viewpoint thinks that the influence area of Earthquake-landslide is the area of external boundary institute enclosing region of all landslides point, and thinks that the experienced earthquake motion intensity in these regions is enough to induced landslide and avalanche.Research finds, the influence factor on earthquake-landslide mainly has: seismologic parameter, geological structure, Rock Nature, topography and geomorphology, hydrogeology, mankind's activity etc..The model of probability of happening of Rockfall hazard after prediction earthquake is built based on these influence factors, after can effectively determining region shake, Rockfall hazard danger is distributed, take precautions against natural calamities for commander, disaster relief work provides decision-making foundation, the directive significance that restoration and reconstruction after calamity, planning and economic development are had, also lays a good foundation for the research of further calamity source simultaneously.

High accuracy remote sensing technology and the development and application of GIS technology, provide favourable technical support for the distribution characteristics of Rockfall hazard after shake and Study on influencing factors, for understanding earthquake-mechanism of the occurrence of landslide, carry out risk assessment further and have important function.

From existing research method, the danger of Rockfall hazard can be divided into two kinds: one to be statistical analysis method, and two is slope stability analysis method.

The forecast model that statistical analysis method is based on the principle of mathematical statistics and method and sets up, general process is the governing factor first impact landslide occurred, as geology, inclination angle, features of terrain etc. adopt mathematical statistics method that the generation on landslide is analyzed with the spatial relationship of influence factor, and attempt finding out statistical law, it is finally based on corresponding rule and the danger that following landslide disaster occurs is made prediction judgement.Adopt mathematical statistics method that landslide, the generation of avalanche and the relation of factor of influence after shake are analyzed, find out statistical law, and predict in following dangerous process the most frequently used to method be Logistic regression analysis.Logistic regression analysis is a kind of probabilistic type nonlinear regression model (NLRM), for studying a kind of multivariable technique of relation between classification observed result and some influence factors.Logistic recurrence classification observed result being only had to two classified variables of two states is known as being that binary Logistic returns.In binary Logistic returns, the substantially probability of the dependent variable in regression equation, rather than variable itself.

Binary Logistic regression equation is expressed as follows:

$P=\frac{\exp ({\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_j{x}_j)}{1+\exp ({\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_j{x}_j)}$ Formulas I

In formula, P is dependent variable, is the independent variable factor probability of happening relative to a certain event, and span is [0,1]；X_jBeing the independent variable factor, j is positive integer, is affect the factor that event occurs；β₀... β_jIt is partial regression coefficient, reflects independent variable factor x_jCapability of influence size to P.In the binary Logistic regression equation used in Rockfall hazard liability, hazard assessment, dependent variable is 1,0 variable, represents " slump ", " non-slump " implication, and independent variable is the influence factor that slump occurs.

Zhao Bin is in the article that exercise question is " the Wenchuan earthquake Study of Hazard Evaluation based on GIS ", see [D]. Capital Normal University, 2011, with Wenchuan County for object of study, establish geological disaster sensitivity assessment model based on GIS and Logistic regression model, and attempt having carried out landslide probability zonation based on sensitivity analysis.Zhou Wei is in the article that exercise question is " studying based on the Logistic Bailong River Basin Landslide Hazard Assessment returned with SINMAP model ", see [D]. Lanzhou University, 2012, analyzing on the basis of landslide disaster factor of influence, have chosen 16 landslide contribution factors, such as include elevation, NDVI, the gradient, lithology, river distance, 60 minutes average rainfalls and Land_use change etc., the minor effect factor is 24 hourly average rainfalls and slope aspect etc. respectively, under the support of GIS, Logistic regression model and SINMAP (StabilityIndexMapping) model are applied in the Landslide Hazard Assessment of Bailong River Basin, research finds that Logistic regression model precision is 70.24%, Regional Landslide Hazard Risk Assessment effect is better than SINMAP model.Utilize statistical analysis method Study of Seismic landslide, the danger of collapse hazard is method more general in current earthquake Rockfall hazard hazard assessment, apply respond well, but this method is based on the statistical law of mathematics mostly, ignore the dynamical mechanism of earthquake Rockfall hazard.

Slope stability analysis method is based on region geotechnical property and mechanical analysis, adopts traditional slope stability computation model that the liability of Regional Landslide, danger are predicted.After shake in Rockfall hazard research, slope stability method can analyze Rockfall hazard dynamical mechanism.Nineteen sixty-five Newmark et al. has carried out the quantitative analysis of the physics origin cause of formation for the reservoir dam failure danger that earthquake is likely to cause, it is proposed to a kind of short-cut method predicting the displacement that comes down under geological process, and the stability of side slope is judged by critical acceleration.The Newmark model later stage is continuously available improvement, and is used widely in Earthquake-landslide risk assessment.Such as, after Jibson et al. have studied 1994 Nian Bei ridge earthquakes, Newmark moves Distribution value, and the landslide cataloguing combining reality constructs Earthquake-landslide prediction curve, see JibsonR.W., HarpE.L., MichaelJ.A. " Amethodforproducingdigitalprobabilisticseismiclandslideh azardmaps ", [J] .EngineeringGeology, 2000,58 (3 4): 271-289.Further Jibson et al. have studied Newmark displacement Dn and (1) Critical Seismic acceleration；(2) critical acceleration rate and magnitude；(3) Arias intensity level and critical acceleration；(4) relation between Arias intensity level and four aspects of critical acceleration rate, see JibsonR.W. " Regressionmodelsforestimatingcoseismiclandslidedisplacem ent ", [J] .EngineeringGeology, 2007,91 (2 4): 209-218.Past research have shown that, Newmark displacement computation model is prediction earthquake-landslide more effective method, but its scope of application has certain limitation, is primarily adapted for use in the shallow failure forecast analysis of rock mass on the one hand；On the other hand, owing to model is higher to the required precision of relevant parameter, the forecast analysis in big region therefore it is difficult to apply to.Build additionally, Newmark displacement computation model is based on ideal ramp, the consideration of actual earthquake Rockfall hazard influence factor is also comprehensive not.

In sum, statistical analysis method and Newmark displacement computation model cut both ways in Rockfall hazard risk analysis after shake.In current research, statistical analysis method and using of Newmark model are substantially independent, and prior art is calculated analyzing only with a kind of method therein.

Accordingly, it would be desirable to build a kind of both considered landslide make a difference the factor spatial distribution it is contemplated that Slope Failure mechanics model shake after Rockfall hazard hazard prediction method, to being correctly predicted the danger of Rockfall hazard after shake.

Summary of the invention

In order to solve above-mentioned technical problem, a kind of both considered landslide make a difference the factor spatial distribution it is contemplated that Slope Failure mechanics model earthquake after there is the Forecasting Methodology of Rockfall hazard.

According to an aspect of the present invention, it is provided that the Forecasting Methodology of Rockfall hazard position occurs after a kind of earthquake, and the method comprises the following steps:

Obtain multiple fundamental geological parameters in space to be predicted, multiple geologic parameters, seismologic parameter, the geographical position of at least one known slump point and the geographical position of multiple non-slump point；

Newmark displacement D is calculated based on Newmark displacement model_n；

With at least some of in multiple fundamental geological relevant parameters and Newmark displacement D_nThe binary Logistic regression equation of Rockfall hazard forecast model is built as the independent variable factor；

By described known slump point and described non-slump point geographical position, the described geographically relevant parameter of each position and calculated Newmark displacement D_nAs the known conditions of described binary Logistic regression equation, calculate the partial regression coefficient of respective Variable Factors in described binary Logistic regression equation；

Calculated partial regression coefficient is utilized to build the Rockfall hazard forecast model in this space to be predicted.

Preferably, what described fundamental geological parameter included in the gradient, topographic relief amplitude, fracture belt position, position, river is one or more；It is one or more that described geologic parameter includes in rock mass completeness, rock-mass quality, rock mass physical parameter and rock group intensity；Described seismologic parameter includes earthquake magnitude, the depth of focus and one or more in epicenter coordinate.

Preferably, the binary Logistic regression equation of described forecast model is as follows:

$P=\frac{\exp ({\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_4{x}_4)}{1+\exp ({\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_4{x}_4)}$

Wherein, P is the probability of happening of slump after any position shake in space to be predicted；

x₁..., x₄Respectively ln (this position is far from river distance), ln (this position is from fracture belt distance), ln (topographic relief amplitude) and ln (D_n)。

Preferably, described non-slump point utilizes the createrandompoints instrument in ArcGIS at the outer stochastic generation of actual slump point.

Preferably, described non-slump point scope stochastic generation outside from actual slump point 200m.

Preferably, Newmark displacement D is calculated according to following formula_n,

logD_n=1.521logI_a-1.993logA_c-1.546±0.375

Wherein, I_aFor Arias intensity, m/s；A_cFor critical acceleration.

Preferably, Arias intensity I is calculated according to equation below_a,

$\log I_a=M-2\log \sqrt{r^2+{7.5}^2}-3.99\±0.5$

Wherein, M is moment magnitude, and r is focal length.

Preferably, using the centrage in maximum earthquake intensity region as linear focus, calculate focal length.

Preferably, the method is predicted suitable in the slip mass thickness Rockfall hazard less than 6m.

According to a further aspect in the invention, it is provided that after the earthquake of a kind of Southwest China occur Rockfall hazard position Forecasting Methodology, including:

Obtain multiple fundamental geological parameters in space to be predicted, multiple geologic parameters, seismologic parameter, the geographical position of at least one known slump point and the geographical position of multiple non-slump point；

Newmark displacement D is calculated based on Newmark displacement model_n；

Utilize following formula to calculate position to be predicted in prediction space and the probability of Rockfall hazard occur,

$P=\frac{\exp ({\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_4{x}_4)}{1+\exp ({\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_4{x}_4)}$

In formula, P is the probability of happening of slump after any position shake in space to be predicted；

x₁..., x₄Respectively ln (this position is far from river distance), ln (this position is from fracture belt distance), ln (topographic relief amplitude) and ln (D_n), wherein,

β₀=2.438～1.297；

β₁=-0.798～-0.524；

β₂=-0.431～-0.436；

β₃=1.272～1.134；

β₄=0.318～0.314.

Under the support of GIS technology and SPSS statistical analysis software, after the spatial relationship of avalanche after extracting impact shake, the Related Environmental Factors that landslide disaster occurs and Related Environmental Factors and Rockfall hazard, can primarily determining that, the gradient, topographic relief amplitude, rock and soil properties, weathering Erosion degree, magnitude etc. are all the key factors of avalanche after impact shake, landslide disaster generation.Simultaneously, it is observed that river and fracture belt have very strong control action for the distribution of Rockfall hazard, mainly due to the existence in river and fracture belt is strengthened both sides, slope ground erosion and weathering, Rock And Soil anti-shear ability is declined, thus causing this regional geological environment abnormal fragile, very easily there is avalanche, landslide disaster.

The method according to the invention, on the basis of Rockfall hazard Analysis on Main Influence Factors after shaking, utilizes the Newmark displacement calculation result of Newmark displacement model, utilizes binary Logistic regression analysis to carry out model foundation and probabilistic forecasting.By choosing Dn value, from river distance, set up probabilistic model from fracture belt distance and four independent variable factors of topographic relief amplitude, calculate the partial regression coefficient of each independent variable factor and build Probabilistic Prediction Model, for the forecast analysis of earthquake rear region Rockfall hazard probability of happening.

Respectively there is the M in Sichuan Province China province Wenchuan County on May 12nd, 2008 in the present invention_sWenchuan County after 8.0 grades of special violent earthquakes, Beichuan County and Mianzhu County are example, the probabilistic model that after establishing the prediction shake being preferably applied to region, Southwestern China portion, Rockfall hazard occurs.Being checked by the comparison and ROC with actual slump point, the probabilistic model prediction accuracy that the present invention sets up reaches more than 75%, it was predicted that effective.

Accompanying drawing explanation

Fig. 1 illustrates the flow chart of the Forecasting Methodology according to the present invention；

Fig. 2 illustrates the earthquake intensity scattergram of various embodiments of the present invention relevant range；

Fig. 3 illustrates the engineering rock component cloth of embodiment 1 relevant range；

Fig. 4 illustrates the terrain slope scattergram of embodiment 1 relevant range；

Fig. 5 illustrates the side slope static security coefficient distribution of embodiment 1 relevant range；

Fig. 6 illustrates the side slope critical acceleration distribution of embodiment 1 relevant range；

Fig. 7 illustrates the Arias intensity distributions of embodiment 1 relevant range；

Fig. 8 illustrates the Newmark accumulation displacement D of embodiment 1 relevant range_nDistribution；

Fig. 9 illustrates the ROC curve of the forecast model obtained according to embodiment 1；

Figure 10 illustrates the distribution of disaster probability of happening and the distribution of slump point of the relevant range according to embodiment 1；

Figure 11 illustrates embodiment 1 relevant range Dn and the distribution of slump point position；

Figure 12 illustrates embodiment 2 relevant range Dn and the distribution of slump point position；

Figure 13 illustrates the ROC curve of embodiment 2；

Figure 14 illustrate embodiment 3 ROC curve；

Figure 15 illustrates the distribution of disaster probability of happening and the distribution of slump point of the relevant range of embodiment 3

Figure 16 illustrates the ROC curve of embodiment 4；

Figure 17 illustrates relevant range Dn and the distribution of slump point position of embodiment 5；

Figure 18 illustrates the ROC curve of the forecast model of embodiment 5；

Figure 19 illustrates the distribution of disaster probability of happening and the distribution of slump point of the relevant range of embodiment 5.

Detailed description of the invention

Come a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings and in conjunction with detailed description of the invention.

According to technical scheme, by being combined with river, fracture belt, topographic relief amplitude by Newmark displacement, utilizing binary Logistic regression model to set up probability function, after carrying out region shake, the probability of happening of Rockfall hazard is predicted.

Describe in detail according to the Forecasting Methodology that Rockfall hazard position occurs after the earthquake of the present invention referring to Fig. 1.

Step S110, after the earthquake, obtains it may happen that multiple fundamental geological parameters in space to be predicted of Rockfall hazard, multiple geologic parameters, seismologic parameter, the geographical position of at least one known slump point and the geographical position of multiple non-slump point.

As known to those skilled in the art, impact shake after Rockfall hazard occur because have geographic factor, such as elevation, the gradient, topographic relief amplitude, fracture belt position and position, river etc.；Geologic parameter, such as formation lithology, including rock mass completeness, rock-mass quality, rock mass physical parameter etc.；And various seismologic parameter, such as earthquake magnitude, the depth of focus and epicenter coordinate etc..The data that the earthquake-stricken area space remote sensing of the establishments such as land Resources Department, China Geological Survey Bureau is met an urgent need in the atlas such as investigation can be utilized to show, obtain slump point position latitude and longitude coordinates.ArcGIS can be utilized at the outer multiple non-slump point of stochastic generation of such as 200m of actual slump point, it is determined that non-slump point position latitude and longitude coordinates.Remote sensing images analysis according to relevant area, it is determined that the fundamental geological relevant parameter that each slump point is relevant to fundamental geological parameter with non-slump point, for instance the elevation of each position, the gradient, topographic relief amplitude, formation lithology, from fracture belt distance and from river distance etc..

Newmark simplifies the seismologic parameter related in displacement computation model and is mainly magnitude, epicentral location and the depth of focus.Seismologic parameter can be obtained according to the basic seismologic parameter of the earthquake that State Seismological Bureau issues and earthquake intensity figure.

Step 120, calculates Newmark displacement D based on Newmark displacement model_n。

Newmark simplify in displacement consider because have the gradient, formation lithology and earthquake intensity.The displacement D of Newmark model_nThe influence factor that after can be regarded as shake, Rockfall hazard occurs, it combines the impact of terrain slope, formation lithology and the Earthquake Intensity factor, and the probability that Rockfall hazard occurs is with D_nIncrease and increase.

It is slope limit equilibrium theory that Newmark displacement calculates the theoretical basis of model construction, it is adaptable to the shallow failure research of earthquake-induced.This model analyzes slope stability based on areal geology, geomorphological environment, thus obtaining the critical acceleration that area slope is subjected to displacement, when an earthquake occurs, after the acceleration that area slope stress produces exceedes critical acceleration, slope will lose stable gradually, slide along surface of fracture, produce permanent displacement, permanent displacement value come down after characterizing shake to a certain extent occur probability size.Here permanent displacement value is to obtain by the difference portion of earthquake external force acceleration Yu critical acceleration is carried out quadratic integral, and its formula can be expressed as:

D_n=∫_t∫_t[a(t)-a_c] dt Formulas I-1

In Formulas I-1, Newmark permanent displacement amount D_nFactor of determination be Earthquake Intensity and critical acceleration a_c.For critical acceleration a_cCalculating, it is common that utilize infinite slope method calculate safety coefficient (F_s), more indirectly solve critical acceleration a_c, calculate process as follows:

Calculate slope static security coefficient F_s:

Formulas I-2

The effective cohesiveness of C', the z surface of fracture degree of depth, m；

γ Rock And Soil severe, N/m³；Z_wGroundwater level depth more than surface of fracture,

m——z_w/ z, dimensionless；β side slope surface inclination angle, (°)；

The severe of γ water, N/m³；Effective angle of inner friction, (°)；

Calculate critical acceleration a_c:

a_c=(F_s-1) gsin β Formulas I-3

Wherein, F_sFor static security coefficient, g is acceleration of gravity, and β is side slope surface inclination angle.

Earthquake Intensity in Formulas I-1 generally utilizes earthquake motion peak acceleration (PGA) and peak velocity (PGV) to describe, and both of which depends on the pulse in short-term of STRONG MOTION DATA medium-high frequency.But there are relation the destructive power of earthquake and frequency of vibration, amplitude, persistent period, frequency of vibration is only used can not completely to reflect the destruction result of earthquake.

1970, American scientist Arias proposed the amount of a comprehensive measurement Earthquake Intensity, i.e. Arias intensity (AriasIntensity), includes frequency of vibration, amplitude and persistent period full detail, more comprehensively reflects the overall condition of earthquake motion.Arias intensity by STRONG MOTION DATA earthquake ground motion acceleration square in group strong earthquakes to time integral, then multiplication by constants is determined, sees Formulas I-4:

$I_a=\frac{\mathrm{\π}}{2g}\underset{0}{\overset{T_d}{\∫}}{\left[a\left(t\right)\right]}^2\mathrm{dt}$ Formulas I-4

Wherein, I_aFor Arias intensity, unit is m/s；A (t) is strong-motion instrument label component acceleration time series；T_dIt it is the persistent period of strong-motion instrument record；T is the time in seconds；G is acceleration of gravity.

Arias intensity is introduced in the research on landslide of Earthquake triggering by Wilson and Keefer et al. at first, utilizes repeatedly STRONG MOTION DATA to obtain the Arias intensity attenuation relation with earthquake magnitude, distance.Wilson et al. utilizes available records data afterwards, by numerical analysis method, improves the region decay empirical equation of Arias intensity, obtains Formulas I-5.

$\log I_a=M-2\log \sqrt{r^2+{7.5}^2}-3.99\±0.5$ Formulas I-5

Wherein, I_aFor Arias intensity, M is moment magnitude, and r is focal length.

Owing to Arias intensity can completely describe Earthquake Intensity, therefore it is introduced in Newmark displacement computation model, in conjunction with critical acceleration a_c, set up corresponding functional equation prediction Newmark displacement D_n, thus for the risk assessment on regional earthquake landslide.

Jibson, et al. R.W. it is being " AMethodforProducingDigitalProbabilisticSeismicLandslideH azardMaps:AnExamplefromtheLosAngeles at exercise question respectively, California, Area ", Technicalreport [R] .USGeologicalSurveyOpen-Filereport, 1997:98-113, with exercise question it is " EvaluatingEarthquake-TriggeredLandslideHazardattheBasinS caleThroughGisintheUpperSeleRiverValley ", [J] .SurveysinGeophysics, 2002, the following logarithmic regression equation Formulas I-6 set up in the article of 23:595-625 is most widely used:

logD_n=1.521logI_a-1.993logA_c-1.546 ± 0.375 Formulas I-6

Wherein, D_nFor Newmark shift value, cm；I_aFor Arias intensity, m/s；A_cFor critical acceleration.

The ultimate principle of Newmark displacement computation model is conventionally known to one of skill in the art, and the requirement of seismologic parameter is reduced by the Newmark model after the simplification shown in above-mentioned formula so that its application is more prone to.

Step S130, with multiple fundamental geological relevant parameters and Newmark displacement D_nThe binary Logistic regression equation I-7 of Rockfall hazard forecast model is built as the independent variable factor,

$P=\frac{\exp ({\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_4{x}_4)}{1+\exp ({\mathrm{\β}}_0+{\mathrm{\β}}_1{x}_1+...+{\mathrm{\β}}_4{x}_4)}$ Formulas I-7

In formula, P is the probability of happening on avalanche after any position shake, landslide in space to be predicted；

x₁..., x₄Respectively ln (this position is far from river distance), ln (this position is from fracture belt distance), ln (topographic relief amplitude) and ln (D_n)。

As previously described, because the influence factor that fail in Newmark model to consider shake comprehensively after, Rockfall hazard occurs, utilize Logistic regression model in this step, at calculated Newmark displacement D_nOn basis, supplement factor of influence, set up Rockfall hazard forecast model after shaking.In the present invention, the factor of influence supplemented needs the predicted position distance from river, from fracture belt distance and topographic relief amplitude.

Step S140, utilizes known conditions to calculate the partial regression coefficient of each Variable Factors in described binary Logistic regression equation.

In this step, for the slump point obtained, can be 1 by dependent variable assignment in binary Logistic regression analysis.For non-slump point, can be 0 by dependent variable assignment in binary Logistic regression analysis.Based on the coordinate of slump point and non-slump point, topographic relief amplitude on respective point position, from fracture belt distance, from river distance and Newmark displacement D_nDeng the numerical value of the independent variable factor, the partial regression coefficient of the independent variable factor can be calculated.

Step S150, utilizes calculated partial regression coefficient to build the Rockfall hazard forecast model in this Rockfall hazard space to be predicted.

The coefficient obtained in step S140 is substituted into equation I-7, obtains the Rockfall hazard forecast model according to the present invention.

A useful aspect according to the present invention, relevant geologic parameter and seismologic parameter can be obtained from the earthquake having occurred and that and the statistical data of Rockfall hazard and build forecast model, the forecast model obtained is applied to and the region of the region adjacent that earthquake occurs or have similar geographical feature area shake after Rockfall hazard predict.Another useful aspect according to the present invention; the method that can utilize the present invention obtains relevant geologic parameter and seismologic parameter from occurent earthquake is real-time with Rockfall hazard and builds forecast model; and utilize constructed forecast model that the position of Rockfall hazard after the contingent shake of this earthquake is predicted, effectively to protect country and the safety of individual's lives and properties.

Illustrate according to the Forecasting Methodology of Rockfall hazard after the earthquake of the present invention below in conjunction with preferred embodiments thereof body.

Embodiment 1

For Rockfall hazard analysis after China's Wenchuan County in Sichuan Province shake, will illustrate according to the forecast model construction method that Rockfall hazard occurs after the earthquake of the present invention below.

Show according to the data in " the emergent investigation of Wenchuan earthquake disaster area space remote sensing " atlas that Ministry of Land and Resources, China Geological Survey Bureau work out, after the shake of Wenchuan County, actual slump point has 1904, the dependent variable assignment of these points is 1 by binary Logistic regression analysis, represents slump point.For non-slump point, this embodiment utilizes the createrandompoints instrument stochastic generation in ArcGIS.Considering the uncertainty that actual slump point is distributed in certain area coverage, embodiment is such as with actual slump point for the center of circle, and 200m is the occurrence scope that radius marks actual slump point, it is assumed that for there is not the region on avalanche, landslide outside this scope.At this, 2000 some positions of stochastic generation in region on avalanche, landslide not occurring, assignment is 0, represents non-slump point.

First, based on slump point as above and non-slump point, extract the coordinate on respective point position, topographic relief amplitude, from fracture belt distance, from river distance, and be used for calculating Newmark displacement D_nVarious geography information and seismologic parameter, for regression analysis.Wenchuan County River Data is to utilize ArcGIS herein, the water system sediments data extracted after carrying out hydrological analysis based on region dem data.

Newmark displacement D is determined subsequently, based on Newmark displacement computation model_n。

1. seismologic parameter

Newmark simplifies the seismologic parameter related in displacement computation model and is mainly magnitude, epicentral location and the depth of focus.After 5.12 Wenchuan violent earthquakes, China Seismology Bureau and country of US Geological Survey earthquake information center have all been issued Wenchuan earthquake and have been included basic seismologic parameter and the earthquake intensity figure of earthquake magnitude, time, the depth of focus and epicenter coordinate etc., see Fig. 2.Owing to, in Newmark simplified model, the parameter in Arias strength calculation formula I-5 is moment magnitude, therefore the magnitude parameter in the present embodiment adopts 7.9 grades of US Geological Survey, and other seismologic parameter adopts the result that China Seismology Bureau announces.

To find out in terms of Fig. 2, Wenchuan earthquake earthquake centre is positioned at Ying Xiu town, depth of focus 14km(China Seismology Bureau), break east Directional Extension northwards.Further, from earthquake centre, northwards east is to extension in the meizoseismal area of Ⅺ earthquake intensity, and area is relatively big, only describes the earthquake impact on surface rupture by focal length distance and is not inconsistent with reality.Therefore, the present embodiment is considered as centrage that earthquake intensity is Ⅺ degree of region as linear focus, and the depth of focus adopts 14km, calculates focal length r.

2. engineering geology parameters

In Newmark simplified model, participate in determining critical acceleration a_cImportant parameter slope static security coefficient F_sCalculate and need according to local landform and lithology data.Wherein lithology data, adopts engineering rock group data to replace in the present embodiment.Engineering rock group data are made up of engineering rock component Butut and corresponding physical and mechanical parameter.The present embodiment determines Wenchuan County engineering rock component Butut with reference to the Wenchuan earthquake severely afflicated area engineering petrofabric diagram of Qi Shengwen et al. compilation, see Fig. 3, referring to Wenchuan earthquake pole severely afflicated area geologic setting and secondary Slope disaster space law of development [J]. engineering geology journal, 2009, (1): 39-49.

Corresponding engineering rock group physical and mechanical parameter, such as effective angle of inner frictionThe Primary Reference " Standard for classification of engineering rock masses " (GB50218-94) such as effective cohesiveness (C'), severe (γ)；" Code for design of building " (GB50007-2002)；" Code for investigation of geotechnical engineering " (GB50021-2001), the standard such as table 1-1,1-2,1-3 is determined.Due to the actual rock mass discontinuity that represents of the potential water use in Newmark displacement model, and structural plane largely controls rock mass strength, and therefore the present embodiment uses the empirical parameter of structural plane to be calculated analyzing.

Showing according to field investigation result, the rock side slope that the hard rock in northeast, Wenchuan County is constituted generally grows 3～4 groups of cracks, rock crushing, integrity is poor, according to table 1-1,1-2,1-3, adjusts accordingly its rock parameter, and parameter corresponding to Wenchuan County each engineering rock group is such as shown in table 1-4.

The qualitative division of table 1-1 rock mass completeness

The classification of table 1-2 rock mass basic quality

Table 1-3 rock mass and structural plane physical and mechanical parameter

Table 1-4 Wenchuan County engineering rock group physical and mechanical parameter

Lithology	Lithology code	Severe (N/m3)	Internal friction angle (°)	Cohesive strength (Pa)
					Hard rock	Ⅰ	27500	29	120000
Relatively hard rock	Ⅱ	26500	24	100000
					Relatively soft rock	Ⅲ	25500	19	80000
Soft rock	Ⅳ	23500	16	70000
					Ultimate soft rock	Ⅴ	21500	13	50000

Table 1-5 Wenchuan County engineering geology rock group strength reduction factor

In addition, it is contemplated that rock mass meets water correction, the effective cohesiveness of its intensity index and internal friction angle are actually low than the empirical value in table 1-4, and therefore the empirical value in his-and-hers watches 1-4 carries out reduction respectively, and reduction relation is table 1-5.

3. other parameter

Determine side slope critical acceleration a_cValue, depend primarily on areal geology, landforms and hydrological environment.From above-mentioned Formulas I-2, Formulas I-3 expression formula it can be seen that domatic inclination angle, ground shear strength, the water-bearing layer degree of depth and the landslide surface degree of depth decide a_cSize.Wherein, region ground shear strength parameter is discussed in engineering geology parameters one saves, and mainly solves other parameters remaining here.

Side slope surface inclination angle (β) is calculated by region DEM altitude data and obtains.Dem data derives from the digital elevation data product of the 30m resolution that national science data service platform is issued herein, Wenchuan County DEM elevation scattergram is obtained by data splicing, fusion, cutting, thus calculating the terrain slope scattergram (30m × 30m) obtaining the whole county, see Fig. 4.Major part region, Wenchuan County terrain slope is more than 30 °, and the slope distribution area less than 10 ° is less, is concentrated mainly on the Xuan Kou town of the southeast, water mill town one band.

The landslide surface degree of depth in Newmark simplified model is landslide thickness, different types of landslide, and its slip mass thickness is different.After Wenchuan earthquake shake, SURVEYING OF LANDSLIDE analytical data shows, the landslide degree of depth in area, Wenchuan is generally less than 3m.According in the Classification of Landslides that landslide thickness carries out, < 6m's is shallow failure to slip mass thickness, and therefore, the landslide after the shake of Wenchuan principally falls into shallow failure, just meets Newmark displacement computation model and is applicable to the optimum condition of shallow failure.But for region, landslide thickness yet suffers from uncertainty, Khazai et al. proposes grade of side slope and the relation of sliding broken thickness, landslide thickness reduces with the increase of the gradient, sees Khazai, B., Sitar, N.Landslidinginnativeground:aGIS-basedapproachtoregional seismicslopestabilityassessment, reporthttp:// www.ce.berkeley.edu/*khazai/Research/, 2000.The present embodiment determines landslide thickness according to this relation, in Table 1-6.

Landslide, table 1-6 Wenchuan thickness and gradient relation

Grade of side slope	Landslide thickness
		0-30°	4m
30-40°	3m
		40-60°	2m
>60°	1m

The water-bearing layer thickness of slip mass is also the parameter considered in calculating.Wenchuan County is located in subtropical zone moist climate band, has a humid climate, and summer is affected by SE Monsoon and southwest monsoon, is in again on windward slope in addition, and high temperature and rainy, annual precipitation is 700～1200mm, and is concentrated mainly on 5～JIUYUE.Therefore, consider the situation that rainfall infiltration causes slip mass fully saturated herein, take the saturated slip mass ratio m=1 to integral thickness.

4. based on Newmark displacement D after the shake of Newmark model_n

Based on ArcGIS software platform, Newmark simplified model is utilized to be capable of the evaluation of Rockfall hazard space liability after shake.According to above-mentioned Formulas I-2, I-3, I-5 and Formulas I-6, according to region geotechnical property and grade of side slope, avalanche, landslide liability and the critical acceleration a that analysis of slope causes due to self build-in attribute_c；Calculate size and the Arias intensity of Earthquake Intensity suffered by different regions；Based on calculated critical acceleration and Arias intensity, calculate Newmark displacement D_n。

(1) critical acceleration a_c

Based on region rock-soil mechanics intensity and regional slope data, adopting formula I-2, I-3 to calculate static security coefficient and critical acceleration, result is as shown in Figure 5 and Figure 6.Critical acceleration scattergram characterizes under identical seismic dynamic loading background, due to the slump easy-suffering level difference that side slope intrinsic property causes.Being left out the earthquake motion impact on zones of different, the region that critical acceleration is more little is more susceptible to Rockfall hazard.

(2) Arias intensity I_a

Calculate Arias intensity according to above-mentioned Formulas I-5, reflect the earthquake motion size that affects on zones of different, shown in result Fig. 7.Owing to focal length herein is to calculate according to the distance between center line from Ⅺ degree of intensity area, therefore, the decline trend of Arias intensity is similar to earthquake intensity decline mode.

(3) Newmark displacement D_n

Based on above-mentioned critical acceleration a_cWith Arias intensity I_aResult of calculation, utilizes above-mentioned Formulas I-6 to calculate the Newmark displacement D obtaining study area_nScattergram, is shown in Fig. 8.

Subsequently, with multiple fundamental geological relevant parameters and calculated Newmark displacement D_nThe binary Logistic regression equation of Rockfall hazard forecast model is built as the independent variable factor.

The present embodiment is on the basis of Newmark model, consider that the factors affect such as river, fracture belt and landform the distribution of Rockfall hazard, Newmark displacement is combined with river, fracture belt and topographic relief amplitude, binary Logistic regression model is utilized to set up probability function, after carrying out region shake, the probability of happening of Rockfall hazard is predicted, sees Formulas I-7.

Each partial regression coefficient β in the Logistic regression equation of table 2-1 Wenchuan County

Variable	Factor beta	Standard error	Waldχ2	Degree of freedom	Significant level	Exp(B)
							Constant	2.438	0.714	11.669	1	0.001	11.456
Ln (from river distance)	-0.798	0.035	505.316	1	0.000	0.450
							Ln (from fracture belt distance)	-0.431	0.035	149.063	1	0.000	0.650
Ln (topographic relief amplitude)	1.272	0.139	83.786	1	0.000	3.568
							ln(Dn)	0.318	0.040	64.413	1	0.000	1.375

Subsequently, by 1904 the slump points obtained in first step and 2000 slump points from river distance, from fracture belt distance, topographic relief amplitude and calculated D_nValue substitutes into above-mentioned Formulas I-7, solves each partial regression coefficient β that equation Formulas I-7 obtains in the Logistic regression equation of Wenchuan County, as shown in table 2-1.

Each partial regression coefficient in table 2-1 is substituted into Formulas I-7 and obtains the occurrence Probability Model Formulas I-8 of Rockfall hazard after mountain area, Wenchuan County shakes:

$P=\frac{\exp (2.438-.0.798x_1-0.431x_2+1.272x_3+0.318x_4)}{1+\exp (2.438-0.798.x_1-0.431x_2+1.{272x}_3+0.318x_4)}$ Formulas I-8

In formula, P is the probability of happening of slump after any position shake in space to be predicted；

x₁..., x₄Respectively ln (this position is far from river distance), ln (this position is from fracture belt distance), ln (topographic relief amplitude) and ln (D_n)。

Partial regression coefficient from table 2-1 it can be seen that slump probability of happening with from river distance, from fracture belt distance become negative correlativing relation, with Newmark displacement D_nBecome positive correlation.Exp (B) is odds ratio, represents that independent variable often changes a unit, and the probability that slump occurs is the multiple of corresponding ratio before change to the ratio occurring without probability.Exp (B) reflects the strength of association of influence factor and Rockfall hazard, as Exp (B) > 1, the risk factor of influence factor and Rockfall hazard increases, positive correlation；Exp (B) < 1, the risk factor of influence factor and Rockfall hazard reduces, negative correlation；Exp (B)=1, influence factor is unrelated with the risk factor of Rockfall hazard.By table 2-1 it can be seen that D_nBeing worth more big, more little from river distance, more little from fracture belt distance region is more susceptible to landslide.

Forecast model fitting effect is checked

As it has been described above, establish the forecast model Formulas I-8 of Rockfall hazard after Wenchuan County shakes by binary Logistic regression analysis.The present embodiment verifies the accuracy of model by ROC curve.

ROC curve is the abbreviation of Receiver Operating Characteristics (ReceiverOperatingCharacteristic), and it is a kind of wide variety of data statistical approach.Application ROC curve can help researcher to determine rational probabilistic classification point, is judged as result (or result does not occur) more than (or less than) how many object of study by probability.During forecast result of model the best, ROC curve should be from lower left corner vertical ascent to top, and then horizontal direction extends right up to the most upper right corner.Fig. 9 is the ROC curve that after the shake of Wenchuan County, Rockfall hazard occurrence Probability Model Formulas I-8 is applied to area, Wenchuan, and its form substantially conforms to the ROC curve form of optimum prediction effect, it is seen then that this model is better to the prediction effect in area, Wenchuan.

Generally, by calculating area under ROC curve (AreaUnderCurve is called for short AUC), quantificational expression model prediction success rate is carried out.Fig. 9 is carried out AUC calculating, and in Table 2-2, the success rate prediction that the result obtained is the Logistic regression model that embodiment 1 is set up reaches 85.3%, it was predicted that respond well, it is possible to for Rockfall hazard probability of happening prediction after actual shake.

Table 2-2AUC statistical analysis

Figure 10 gives size and actual avalanche, the landslide disaster point position distribution situation of avalanche after the shake of Rockfall hazard occurrence Probability Model Formulas I-8 prediction, landslide disaster probability of happening after the shake set up according to embodiment 1.From this figure, it can be seen that the high region of Rockfall hazard probability of happening is relatively concentrated after shake, mainly along river, fracture belt be that ribbon is distributed.All in all, east and in North zone the area in Rockfall hazard region occurred frequently relatively big, be mainly distributed on these areas mainly due to tectonic faults, simultaneously Newmark shift value D_nAlso higher in eastern region, these combined factors get up to control macroscopical distribution of Rockfall hazard.The probabilistic forecasting result occurred from Rockfall hazard and the contrast effect of actual slump point position, the high-risk danger zone of Rockfall hazard is more consistent with the distribution ratio of actual Rockfall hazard.

Comparative example 1

Figure 11 gives Wenchuan County Newmark shift value D_nDistribution and actual avalanche, landslide disaster point position distribution situation.From this figure, it can be seen that D_n> distributed areas of region and actual slump point position of 3cm generally compare and coincide, but by extracting the D of slump point position respective point_nIt is worth, and statistics drops on D_n> actual slump point number in 3cm region finds, D_n> actual slump incidence rate in 3cm region is only about 35%, it is seen that Newmark model is better to the prediction of Rockfall hazard distribution trend, but the accuracy that single Rockfall hazard is predicted is relatively low.

Newmark simplifies displacement model and depends on side slope lithologic character and the difference of earthquake motion impact, although more comprehensively consider avalanche, come down the engineering lithologic character and triggering factors that occur, but have ignored region landform, physiognomy pattern feature.Such as, lack and river dissection is considered, lack the consideration etc. on rift structure impact.

Relatively comparative example 1 can be seen that, according to Rockfall hazard occurrence Probability Model I-8 after the shake set up based on Newmark Displacement Analysis and binary Logistic regression analysis of the present invention, the predictablity rate of large-scale Lan-cang River and local danger position is high, prediction effect is good, it is possible to take into account two aspects of both macro and micro.

Embodiment 2

Prediction effect when this embodiment verifies that the forecast model Formulas I-8 that embodiment 1 is set up is applied to Beichuan County, Sichuan Province by ROC curve,.

Beichuanqiangzu Autonomous County is positioned at In Northwest of Sichuan Basin.Geographical coordinate: north latitude 31 ° 35 ' 31 ° 38 ' 2 ", east longitude 104 ° 26 ' 15 " 104 ° 29 ' 10 ".East connects Jiangyou City, adjacent An County, south, west depends on Mao County, and Songpan, Pingwu are supported in north, and territory area is 2867.83 sq-kms.

According to Rockfall hazard occurrence Probability Model Formulas I-8 after the shake that inventive embodiments 1 is set up, Beichuan County regional earthquake is brought out avalanche, the probability of happening prediction process of landslide disaster is broadly divided into three parts: obtains geographic factor, geologic parameter and seismologic parameter, calculates the Newmark displacement D in this region_nApplication effect with the forecast model evaluating the method according to the invention and set up and set up model.

First, the various relevant parameters of Beichuan County are obtained according to step as described in Example 1.

Subsequently, acquired geologic parameter and seismologic parameter is utilized to calculate Newmark displacement D_n。

Beichuan County slope map, based on the DEM of 30m × 30m, is obtained by the ratio of the depth displacement between calculating adjacent cells unit with horizontal range.Relevant parameter is brought into Formulas I-2, Formulas I-3, Formulas I-5 and Formulas I-6 calculate the Newmark accumulation shift value D of each grid point value in Beichuan County_n, the liability of Rockfall hazard after analyzed area shake accordingly.Figure 12 is calculated Beichuan County Newmark accumulative displacement distribution situation.

Subsequently, calculating is utilized to obtain the distribution of Newmark accumulative displacement and the Beichuan County geographic factor obtained, it is possible to the forecast model Formulas I-8 set up in embodiment 1 is used for the prediction of Rockfall hazard probability of happening after Beichuan County is shaken.

In order to check forecast model in the application effect of Beichuan County, the present embodiment, by the ArcGIS outer non-slump point of stochastic generation 617 of 200m scope at 617 actual slump points of Beichuan County, analyzes this model prediction effect at Beichuan.

Equally, ROC curve is adopted to carry out testing model prediction effect.Figure 13 is that the forecast model Formulas I-8 of embodiment 1 foundation ROC after Beichuan County is applied checks curve.From tracing pattern, the probability of happening of Rockfall hazard after the shake of Beichuan County is had good predictive ability by the forecast model of embodiment 1 equally.From AUC result of calculation, in Table 2-3, the success rate prediction of Rockfall hazard after the shake of Beichuan County is about 80.3% by the forecast model that embodiment 1 is set up.

Table 2-3AUC statistical analysis

In sum, the forecast model that embodiment 1 is set up is better to the prediction effect of the avalanche of Beichuan County earthquake-induced, Landslide hazards.This model is based on the forecast model that the actual slump point in Wenchuan is set up, show that this model has for similar area by this embodiment application verification in Beichuan County and extend the suitability preferably, can be used in the quantification risk assessment of the earthquake Rockfall hazard in Southwestern China mountain area, lay the foundation for the preventing and treating of Secondary Geological Hazards after shake.

Embodiment 3

Rockfall hazard analysis after shaking for Beichuan County, Sichuan Province of China below, illustrates according to the forecast model construction method that Rockfall hazard occurs after the earthquake of the present invention.

First, obtain the various relevant parameter in Beichuan County and determine Newmark displacement Dn based on Newmark displacement computation model.

Concrete steps are referring to embodiment 2 related content, and result of calculation is as shown in figure 12.

Subsequently, based on occur 617 actual slump points of Beichuan and in Beichuan County 617 non-slump points of the outer stochastic generation of the 200m scope of actual slump point, dependent variable assignment to slump point is 1, dependent variable assignment to non-slump point assignment is 0, utilize acquired various geographic factors and calculated Newmark displacement Dn etc. as known conditions, binary Logistic regression equation I-7 is solved, obtain each partial regression coefficient β in the Logistic regression equation of Beichuan County, as shown in Table 2-4.

Each partial regression coefficient β in the Logistic regression equation of table 2-4 Beichuan County

Variable	Coefficient	Standard error	Waldχ2	Degree of freedom	Significant level	Exp(B)
							Constant	1.297	0.689	3.541	1	0.060	3.658
Ln (from river distance)	-0.524	0.032	271.904	1	0.000	0.592
							Ln (from fracture belt distance)	-0.436	0.034	166.075	1	0.000	0.646
Ln (topographic relief amplitude)	1.134	0.133	72.381	1	0.000	3.107
							ln(Dn)	0.314	0.038	68.762	1	0.000	1.368

Partial regression coefficient in table 2-4 is substituted into Formulas I-7 and obtains the probability of happening forecast model Formulas I-9 of Rockfall hazard after mountain area, Beichuan County shakes:

$P=\frac{\exp (1.297-.0.524x_1-0.436x_2+1.134x_3+0.314x_4)}{1+\exp (1.297-0.524.x_1-0.436x_2+1.{134x}_3+0.314x_4)}$ Formulas I-9

In formula, P is the probability of happening on avalanche after any position shake, landslide in space to be predicted；

x₁..., x₄Respectively ln (this position is far from river distance), ln (this position is from fracture belt distance), ln (topographic relief amplitude) and ln (D_n)。

Figure 14 illustrate shake after the ROC curve being applied to Beichuan County of Rockfall hazard occurrence Probability Model I-9.By calculating area AUC under ROC curve, in Table 2-5, carry out quantificational expression model prediction success rate, the success rate that the result obtained is the prediction Beichuan Rockfall hazard generation of the Logistic regression model Formulas I-9 that embodiment 3 is set up reaches 84.5%, prediction effect is good, it is possible to for Rockfall hazard probability of happening prediction after actual shake.

Table 2-5AUC statistical analysis

Figure 15 give the prediction shake set up according to embodiment 3 after avalanche after the shake of probabilistic model Formulas I-9 prediction that occurs of Rockfall hazard, landslide disaster probability of happening size and actual avalanche, landslide disaster point position distribution situation.From this figure, it can be seen that the high region of Rockfall hazard probability of happening is relatively concentrated after shake, mainly along river, fracture belt, road be that ribbon is distributed.The probabilistic forecasting result occurred from Rockfall hazard and the contrast effect of actual slump point position, the high-risk danger zone of Rockfall hazard is more consistent with the distribution ratio of actual Rockfall hazard.

Embodiment 4

This embodiment verifies the prediction effect of-9 pairs of Wenchuan County of forecast model Formulas I that embodiment 3 sets up by two indexs of ROC curve.

According to Rockfall hazard occurrence Probability Model Formulas I-9 after the shake that inventive embodiments 3 is set up, Wenchuan County regional earthquake is brought out avalanche, the probability of happening prediction process of landslide disaster is broadly divided into three parts: obtain geographic factor, geologic parameter and seismologic parameter；Calculate the Newmark displacement Dn in this region and evaluate the disaster occurrence Probability Model applied.

Obtaining the relevant geographic factor in Wenchuan County, geologic parameter, the related content of seismologic parameter and calculating Newmark displacement Dn has specifically described in embodiment 1, repeats no more here.Each relevant parameter of obtained Wenchuan is substituted into forecast model Formulas I-9 set up in embodiment 3, can be used for the prediction of slump probability after Wenchuan County is shaken.

Table 2-6AUC statistical analysis

Figure 16 is that after shake, the Rockfall hazard probability of happening forecast model Formulas I-9 ROC after Wenchuan County is applied checks curve, from tracing pattern, it was predicted that the probability of happening of Rockfall hazard after the shake of Wenchuan County is had good predictive ability by modular form I-9.From AUC result of calculation, in Table 2-6, after-9 pairs of Wenchuan County shakes of the forecast model Formulas I of embodiment 3, the success rate prediction of Rockfall hazard is about 82.9%, it was predicted that effect is good equally.

Embodiment 5

Rockfall hazard analysis after shaking for Mianzhu City of Sichuan Province of China below, illustrates according to the forecast model construction method that Rockfall hazard occurs after the earthquake of the present invention.

First, obtaining the various relevant parameter in Mianzhu City and determine Newmark displacement Dn based on Newmark displacement computation model, result of calculation is as shown in figure 17.

Subsequently, based on occur 281 actual slump points of Mianzhu City and in Mianzhu City 281 non-slump points of the outer stochastic generation of the 200m scope of actual slump point, dependent variable assignment to slump point is 1, dependent variable assignment to non-slump point assignment is 0, utilize binary Logistic regression equation E, obtain each partial regression coefficient β in Mianzhu City's Logistic regression equation, as shown in table 2-7.

Partial regression coefficient in table 2-7 is substituted into Formulas I-7 and obtains the probability of happening forecast model Formulas I-10 of Rockfall hazard after mountain area, Mianzhu City shakes:

Each partial regression coefficient β in table 2-7 Mianzhu City Logistic regression equation

Variable	Coefficient	Standard error	Waldχ2	Degree of freedom	Significant level	Exp(B)
							Constant	-1.534	1.724	0.792	1	0.374	0.182
Ln (from river distance)	-1.633	0.173	88.877	1	0.000	0.195
							Ln (from fracture belt distance)	-0.056	0.061	0.840	1	0.360	0.945
Ln (topographic relief amplitude)	2.722	0.321	72.151	1	0.000	15.218
							ln(Dn)	0.394	0.131	9.091	1	0.000	1.482

$P=\frac{\exp (-1.534-1.633x_1-0.056x_2+2.722x_3+0.394x_4)}{1+\exp (-1.534-1.633x_1-0.056x_2+2.722x_3+0.394x_4)}$ I-10

In formula, P is the probability of happening on avalanche after any position shake, landslide in space to be predicted；

x₁..., x₄Respectively ln (this position is far from river distance), ln (this position is from fracture belt distance), ln (topographic relief amplitude) and ln (D_n)。

After Figure 18 illustrates shake, Rockfall hazard occurrence Probability Model Formulas I-10 is applied to the ROC curve of Mianzhu City.It is called for short AUC by calculating area under ROC curve, in Table 2-8, carry out quantificational expression model prediction success rate, the success rate that the result obtained is prediction Mianzhu City Rockfall hazard generation of the Logistic regression model that embodiment 5 is set up reaches 92.6%, prediction effect is good, it is possible to for Rockfall hazard probability of happening prediction after actual shake.

Figure 19 give the prediction shake set up according to embodiment 5 after avalanche after the shake of probabilistic model Formulas I-10 prediction that occurs of Rockfall hazard, landslide disaster probability of happening size and actual avalanche, landslide disaster point position distribution situation.From this figure, it can be seen that the high region of Rockfall hazard probability of happening is relatively concentrated after shake, mainly along river, fracture belt, road be that ribbon is distributed.The probabilistic forecasting result occurred from Rockfall hazard and the contrast effect of actual slump point position, the high-risk danger zone of Rockfall hazard is more consistent with the distribution ratio of actual Rockfall hazard.

Table 2-8AUC statistical analysis

Should be appreciated that above is illustrative and not restrictive by preferred embodiment to the detailed description that technical scheme carries out.Technical scheme described in each embodiment can be modified by those of ordinary skill in the art on the basis of reading description of the present invention, or wherein portion of techniques feature carries out equivalent replacement；And these amendments or replacement, do not make the essence of appropriate technical solution depart from the spirit and scope of various embodiments of the present invention technical scheme.

Claims (9)

1. after earthquake there is a Forecasting Methodology for position in Rockfall hazard, and the method comprises the following steps:

Obtain multiple fundamental geological parameters in space to be predicted, multiple geologic parameters, seismologic parameter, the geographical position of at least one known slump point and the geographical position of multiple non-slump point；

Newmark displacement D is calculated based on Newmark displacement model_n；

With at least some of in multiple fundamental geological parameters and Newmark displacement D_nThe binary Logistic regression equation of Rockfall hazard forecast model is built as the independent variable factor；

By described known slump point and described non-slump point geographical position, the described fundamental geological parameter of each position and calculated Newmark displacement D_nAs the known conditions of described binary Logistic regression equation, calculate the partial regression coefficient of respective Variable Factors in described binary Logistic regression equation；

Calculated partial regression coefficient is utilized to build the Rockfall hazard forecast model in this space to be predicted,

It is characterized in that, the binary Logistic regression equation of described forecast model is as follows:

Wherein, P is the probability of happening of slump after any position shake in space to be predicted；

x₁..., x₄Respectively ln (this position is far from river distance), ln (this position is from fracture belt distance), ln (topographic relief amplitude) and ln (D_n),

,,…,The respectively partial regression coefficient of respective Variable Factors.

2. after earthquake as claimed in claim 1 there is the Forecasting Methodology of position in Rockfall hazard, it is characterised in that it is one or more that described fundamental geological parameter includes in the gradient, topographic relief amplitude, fracture belt position, position, river；It is one or more that described geologic parameter includes in rock mass completeness, rock-mass quality, rock mass physical parameter and rock group intensity；Described seismologic parameter includes earthquake magnitude, the depth of focus and one or more in epicenter coordinate.

3. after earthquake as claimed in claim 1 there is the Forecasting Methodology of position in Rockfall hazard, it is characterised in that described non-slump point utilizes the createrandompoints instrument in ArcGIS at the outer stochastic generation of known slump point.

4. after earthquake as claimed in claim 3 there is the Forecasting Methodology of position in Rockfall hazard, it is characterised in that described non-slump point scope stochastic generation outside from known slump point 200m.

5. after earthquake as claimed in claim 1 there is the Forecasting Methodology of position in Rockfall hazard, it is characterised in that calculates Newmark displacement D according to following formula_n,

Wherein, I_aFor Arias intensity, m/s；A_cFor critical acceleration.

6. after earthquake as claimed in claim 5 there is the Forecasting Methodology of position in Rockfall hazard, it is characterised in that

Wherein, M is moment magnitude, and r is focal length.

7. after earthquake as claimed in claim 6 there is the Forecasting Methodology of position in Rockfall hazard, it is characterised in that using the centrage in maximum earthquake intensity region as linear focus, calculates focal length.

8. after earthquake as claimed in claim 5 there is the Forecasting Methodology of position in Rockfall hazard, it is characterised in that the method is applicable to the slip mass thickness Rockfall hazard less than 6m and predicts.

9. after the earthquake of Southwest China there is a Forecasting Methodology for position in Rockfall hazard, including:

Obtain multiple fundamental geological parameters in space to be predicted, multiple geologic parameters, seismologic parameter, the geographical position of at least one known slump point and the geographical position of multiple non-slump point；

Newmark displacement D is calculated based on Newmark displacement model_n；

Following formula is utilized to calculate the probability of happening of position to be predicted Rockfall hazard in prediction space,

P is the probability of happening of slump after any position shake in space to be predicted；

x₁..., x₄Respectively ln (this position is far from river distance), ln (this position is from fracture belt distance), ln (topographic relief amplitude) and ln (D_n),

,,…,The respectively partial regression coefficient of respective Variable Factors,

Wherein,

=2.438 ~ 1.297；

=-0.798 ~-0.524；

=-0.431 ~-0.436；

=1.272 ~ 1.134；

=0.318~0.314。