Regional geological disaster multi-index space risk assessment model and implementation method and application thereof

Abstract

The invention discloses a regional geological disaster multi-index space risk assessment model and an implementation method and application thereof. The implementation method comprises the following steps: S1, collecting and preprocessing geological disaster occurrence data and environmental influence factor data; s2, calculating the contribution degree of each environmental influence factor to the occurrence of the geological disaster, and screening out a main control factor; s3, by taking the geological disaster occurrence data as dependent variables and the main control factors as independent variables, constructing a Bayesian space sharing hyper-parameter model considering geological disaster susceptibility and geological disaster intensity; when the model is used for evaluating the regional geological disaster multi-index risk, the method comprises the following steps: outputting the comparable influence of a shared environment factor on disaster occurrence by using a Bayesian space shared hyper-parameter model, identifying a shared space random effect, and predicting geological disaster susceptibility and intensity. According to the invention, spatial integrated joint risk assessment of the susceptibility and strength of regional geological disasters can be realized, and a more accurate and comprehensive multi-dimensional assessment result can be obtained.

Description

Technical Field

[0001] The present invention relates to the field of geological disaster technology, and in particular to a regional geological disaster multi-index spatial risk assessment model and an implementation method and application thereof. Background Art

[0002] Geological disasters cause a large number of casualties and significant economic losses every year, and their impacts are increasing over time. Comprehensively assessing the risk of regional geological hazards and accurately identifying high-risk areas are of great significance to improving the disaster response capabilities of various regions. In response to this global challenge, implementing entities are encouraged to commit to comprehensive assessments and understand the many different dimensions of disaster risks. More importantly, a better understanding of the interrelated nature of risks is essential to effectively reduce disaster risks and achieve sustainable development goals.

[0003] Mainstream regional geohazard research focuses on assessing susceptibility, that is, assessing the probability of disasters occurring within a spatial unit. However, these studies often ignore the number of disasters occurring within a spatial unit, and the evaluation results cannot accurately distinguish between areas with different disaster intensities (how many disasters occurred or the area where the disaster occurred). Compared with susceptibility assessment, assessing geohazard intensity provides a more comprehensive hazard quantification by identifying areas with multiple disasters and greater potential for destructiveness. In addition, geohazard intensity assessment identifies areas where multiple disasters may occur that are smaller and more dangerous than susceptibility assessments, enabling emergency personnel to accurately focus on and locate key intervention areas.

[0004] Traditional geological hazard risk assessment methods mostly focus on a single dimension, such as geological hazard susceptibility (the probability of a disaster occurring in a certain spatial unit) or geological hazard intensity (the frequency or area of ​​a disaster occurring in a certain spatial unit). Although these methods provide a preliminary basis for the identification of geological hazard risks, in complex areas where disasters occur frequently, this single-dimensional assessment method is difficult to fully reflect the true hazard characteristics of geological hazards. In particular, when multiple geological hazard events occur in the same spatial unit, traditional assessment methods are usually unable to effectively identify these multiple hazards, resulting in one-sidedness and inaccuracy in risk assessment.

[0005] In recent years, some studies have begun to attempt to jointly assess the susceptibility and intensity of geological hazards in an attempt to provide more comprehensive hazard information. However, most of these studies use a separate modeling framework, that is, to independently model susceptibility and intensity. Although this separate framework has enriched the content of disaster risk assessment to a certain extent, the correlation between susceptibility and intensity has not been considered in this framework. Since susceptibility and intensity are affected by the same topographic, climatic and geological conditions, their hazard characteristics are essentially related (non-independent). Separate modeling will lead to a lack of consistency and comparability of the two hazard results, especially when analyzing the joint effect of influencing factors on the two hazards, the results will have certain deviations. Summary of the invention

[0006] In view of the above problems, the present invention aims to provide a regional geological disaster multi-index spatial hazard assessment model and its implementation method and application.

[0007] The technical solution of the present invention is as follows:

[0008] In a first aspect, a method for implementing a regional geological disaster multi-index spatial risk assessment model is provided, comprising the following steps:

[0009] S1: Collect and preprocess geological disaster occurrence data and environmental impact factor data;

[0010] S2: Calculate the contribution of each environmental impact factor to the occurrence of geological disasters, and screen out the main controlling factors according to the contribution;

[0011] S3: Taking the geological disaster occurrence data as the dependent variable and the main controlling factors as the independent variables, a Bayesian spatial shared hyperparameter model is constructed that takes into account the susceptibility and intensity of geological disasters.

[0012] Preferably, in step S1, the environmental impact factors in the environmental impact factor data include earthquake factors, topographic factors, geological and hydrological factors, and human engineering activity factors.

[0013] Preferably, the earthquake triggering factors include peak ground acceleration, peak ground velocity, earthquake intensity, distance from the epicenter of the earthquake, and distance from the fault;

[0014] The topographic factors include altitude, ground roughness, terrain relief, slope, aspect, topography and land cover;

[0015] The geohydrological factors include lithology, vegetation cover index, average annual precipitation, and distance to the nearest river;

[0016] The human engineering activity factors include the distance to the nearest settlement and the distance to the nearest road.

[0017] Preferably, in step S2, a random forest model is used to calculate the contribution of each environmental influencing factor to the occurrence of geological disasters.

[0018] Preferably, in step S2, before calculating the contribution, a collinearity test is also included to ensure that the VIF value is less than a threshold.

[0019] Preferably, in step S3, the Bayesian space sharing hyperparameter model is:

[0020]

[0021] Where: η i,q is the linear predictor of the target variable for each slope unit i in group q; g(·) is the linear predictor of η i,q Likelihood function connected with observed data; g q is the likelihood function of the q group of target variables; Y i (1) Y is the susceptibility to geological disasters; i (2) is the intensity of geological disasters; f(·) is a different potential Gaussian model; α q is the independent intercept term within group q; K is the number of influencing factors; β k,q is the grouped regression coefficient of the kth environmental covariate in group q; X k is the value of the kth influencing factor; ξ i,q is the shared spatial intercept random effect of the target variable of q groups; ε i,q is the residual of each group; i is the spatial intercept random effect for each spatial unit; ξ -i is the adjacent spatial unit of slope unit i; N(·) is a normal distribution; is the weight w in a given space i Under the condition of -i Mean; w i is the spatial adjacency matrix of spatial unit i; δ ξ is the variance of the spatial intercept random effect; is the total number of adjacent spatial units of spatial unit i; ξ j is the random effect value of the spatial intercept of the jth adjacent spatial unit of spatial unit i.

[0022] Preferably, the susceptibility of geological disasters obeys a binomial distribution, and the intensity of geological disasters obeys a Poisson distribution.

[0023] In the second aspect, a regional geological disaster multi-index spatial hazard assessment model is provided, which is established using any of the implementation methods of the regional geological disaster multi-index spatial hazard assessment model described above.

[0024] The third aspect provides the application of the above-mentioned multi-index spatial risk assessment model for regional geological disasters in the multi-index risk assessment of regional geological disasters.

[0025] As a preferred method, the multi-index risk assessment of regional geological hazards specifically includes the following steps:

[0026] Using the Bayesian spatial shared hyperparameter model to output the comparable impacts of shared environmental factors on disaster occurrence;

[0027] identifying shared spatial random effects using the Bayesian spatial shared hyperparameter model;

[0028] The Bayesian spatial sharing hyperparameter model is used to predict the susceptibility and intensity of geological disasters.

[0029] The beneficial effects of the present invention are:

[0030] The present invention can realize spatially integrated joint risk assessment of regional geological disaster susceptibility and intensity by establishing a Bayesian spatial shared hyperparameter model that takes into account the susceptibility and intensity of geological disasters, and obtain more accurate assessment results. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0032] Figure 1 It is a flowchart of the implementation method and application of the regional geological disaster multi-index spatial risk assessment model of the present invention;

[0033] Figure 2 This is a schematic diagram of the contribution of various environmental factors to the occurrence of regional landslides in a specific embodiment; Figure 2 (a) is a schematic diagram of the susceptibility contribution results. Figure 2 (b) is a schematic diagram of the intensity contribution results;

[0034] Figure 3 A schematic diagram of the common dangerous impact of various main controlling factors on the susceptibility and intensity of regional landslides in a specific embodiment;

[0035] Figure 4 It is a schematic diagram of the spatial autocorrelation random effect results of regional landslide in a specific embodiment; wherein, Figure 4 (a) is the random effect of spatial autocorrelation of landslide susceptibility, Figure 4 (b) is the random effect of spatial autocorrelation of landslide intensity;

[0036] Figure 5 A schematic diagram of regional landslide susceptibility hazards in a specific embodiment;

[0037] Figure 6 It is a diagram of absolute intensity and relative intensity of regional landslide in a specific embodiment. DETAILED DESCRIPTION

[0038] The present invention is further described below in conjunction with the accompanying drawings and embodiments. It should be noted that, in the absence of conflict, the embodiments in this application and the technical features in the embodiments can be combined with each other. It should be noted that, unless otherwise specified, all technical and scientific terms used in this application have the same meanings as those generally understood by those of ordinary skill in the art to which this application belongs. The words "including" or "comprising" and the like used in the disclosure of the present invention mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects.

[0039] First, as Figure 1 As shown, the present invention provides a method for implementing a regional geological disaster multi-index spatial risk assessment model, comprising the following steps:

[0040] S1: Collect and preprocess geological disaster occurrence data and environmental impact factor data.

[0041] In a specific embodiment, the environmental impact factors in the environmental impact factor data include earthquake factors, topographic factors, geological and hydrological factors, and human engineering activity factors.

[0042] In a specific embodiment, the earthquake triggering factors include peak ground acceleration, peak ground velocity, earthquake intensity, distance from the epicenter of the earthquake and distance from the fault; the topographic factors include altitude, ground roughness, terrain undulation, slope, slope aspect, topography and land cover; the geological and hydrological factors include lithology, vegetation cover index, average annual precipitation and distance to the nearest river; the human engineering activity factors include the distance to the nearest settlement and the distance to the nearest road.

[0043] S2: Calculate the contribution of each environmental influencing factor to the occurrence of geological disasters, and screen out the main controlling factors based on the contribution.

[0044] Selecting the main control factor to construct the hazard detection index system is an important step in hazard assessment, which is directly related to the data input part of the model. Different combinations of environmental factors and the quality of factor data will affect the final evaluation accuracy. The main factors that lead to geological disasters are varied, but due to the different ecological environments in different research areas, the main control factors will also change. The present invention screens the main control factors by calculating the contribution of each environmental impact factor to the occurrence of geological disasters, and can remove environmental impact factors with lower contribution, avoiding the inclusion of environmental impact factors with lower contribution (i.e., the influence of the explanatory variable on the target variable is too small) in the regression model to affect other relatively important variables.

[0045] In a specific embodiment, the random forest model is used to calculate the contribution of each environmental impact factor to the occurrence of geological disasters. The contribution calculated using the random forest model is mainly measured by the increase in node purity, and the specific calculation formula is:

[0046]

[0047] Where B is the number of trees in the random forest; t is the index of each tree; v is the node corresponding to each tree t; I(v) is an indicator function, which is 1 if node v is split on variable m, otherwise it is 0; Δi(v) is the reduction in impurity due to the split.

[0048] From the perspective of overall stability, the greater the value of the node purity increase, the greater the relative contribution of the explanatory variable to the target variable. In the invention, it means that the influencing factor has a greater impact on the occurrence of disasters.

[0049] In a specific embodiment, before calculating the contribution, a step of performing a collinearity test to ensure that the VIF value is less than a threshold is also included.

[0050] S3: Taking the geological disaster occurrence data as the dependent variable and the main controlling factors as the independent variables, a Bayesian spatial shared hyperparameter model is constructed that takes into account the susceptibility and intensity of geological disasters.

[0051] In a specific embodiment, the Bayesian spatial sharing hyperparameter model is:

[0052]

[0053] Where: η i,q is the linear predictor of the target variable for each slope unit i in group q; g(·) is the linear predictor of η i,q Likelihood function connected with observed data; g q is the likelihood function of the q group of target variables; Y i (1) Y is the susceptibility to geological disasters;i (2) is the intensity of geological disasters; f(·) is a different potential Gaussian model; α q is the independent intercept term within group q; K is the number of influencing factors; β k,q is the grouped regression coefficient of the kth environmental covariate in group q; X k is the value of the kth influencing factor; ξ i,q is the shared spatial intercept random effect of the target variable of q groups; ε i,q is the residual of each group; i is the spatial intercept random effect for each spatial unit; ξ -i is the adjacent spatial unit of slope unit i; N(·) is a normal distribution; is the weight w in a given space i Under the condition of -i Mean; w i is the spatial adjacency matrix of spatial unit i; δ ξ is the variance of the spatial intercept random effect; is the total number of adjacent spatial units of spatial unit i; ξ j is the random effect value of the spatial intercept of the jth adjacent spatial unit of spatial unit i.

[0054] In a specific embodiment, the susceptibility of geological disasters obeys a binomial distribution, and the intensity of geological disasters obeys a Poisson distribution.

[0055] In the above embodiment, the Bayesian spatial shared hyperparameter model uses the Bayesian hierarchical realization method to estimate the shared spatial random effects of geological disaster occurrence and number of geological disasters and the fixed effects of covariates. Through this model, the hazard map of regional geological disaster susceptibility and intensity can be predicted simultaneously, without the need to use two separate models in a separate framework as in traditional methods. The Bayesian spatial shared hyperparameter model is established by the following steps:

[0056] For the susceptibility of geological disasters, each slope unit i has two possibilities: geological disasters occur (1) or do not occur (0). Therefore, it is defined as a binomial distribution: Y i (1) ~Binomial(n i ,P i ), where n = 1, P i Indicates the probability of geological disasters occurring.

[0057] For the intensity of geological disasters, the number of geological disasters Y occurring in each slope unit i is counted. i (2) , which follows a Poisson distribution: Y i (2) ~Poisson(Ei λ i ), where E i represents the expected number of geological disasters in the unit, λ i Relative intensity. A relative intensity greater than 1 indicates that the geological hazard risk in the area is high. Compared with the absolute intensity measurement of only observing the number of geological disasters, estimating relative intensity can provide a more comprehensive understanding of the regional geological hazard risk.

[0058] The Bayesian spatial shared hyperparameter model described in the present invention contains two likelihood functions: one represents the probability of occurrence of geological hazards (susceptibility), and the other represents the number of occurrences of geological hazards (intensity). Therefore, a two-column response matrix is ​​defined, the number of columns corresponds to the number of likelihood functions, and the number of rows is equal to the total number of observations 2i. In addition, for each shared hyperparameter component, two index columns are created: one index ranges from 1 to i and is repeated once to cover two likelihood functions; the first half of the other index column is set to 1 and the second half is set to 2. This setting allows the hyperparameters to integrate all observation information while retaining the independent potential effects of different likelihood functions, thereby retaining more features.

[0059] The Bayesian spatial sharing hyperparameter model described in the present invention divides the target variable into Q groups, represented by q=1,...,Q. In this case, a vector consisting of two positive integers is defined to distinguish between susceptibility and intensity. In addition, according to the first law of geography, the occurrence of geological disaster events is not completely random. Due to the existence of unobserved effects, geological disasters may show geographic spatial clustering in a small area, resulting in spatial autocorrelation. Therefore, the Bayesian spatial sharing hyperparameter model described in the present invention deals with the spatial autocorrelation problem by introducing a conditional autoregressive prior model (Formulas (2)-(3)).

[0060] In a second aspect, the present invention also provides a regional geological disaster multi-index spatial hazard assessment model, which is established using any of the above-mentioned methods for implementing the regional geological disaster multi-index spatial hazard assessment model.

[0061] Thirdly, Figure 1 As shown, the present invention also provides the application of the above-mentioned regional geological disaster multi-index spatial risk assessment model in the regional geological disaster multi-index risk assessment.

[0062] In a specific embodiment, assessing the multi-indicator hazard of regional geological disasters specifically includes the following steps: using the Bayesian spatial shared hyperparameter model to output the comparable impact of shared environmental factors on the occurrence of disasters, identifying shared spatial random effects, and predicting the susceptibility and intensity of geological disasters.

[0063] In a specific embodiment, taking Wenchuan as an example, the regional geological disaster multi-index spatial risk assessment model and its implementation method and application described in the present invention are used to perform a regional geological disaster (landslide) multi-index risk assessment, which specifically includes the following steps:

[0064] (1) Collect and preprocess geological disaster occurrence data and environmental impact factor data;

[0065] The Wenchuan earthquake (May 12, 2008, 8.0 on the Richter scale) triggered large-scale landslide activity, of which a relatively complete list of landslides was determined by Xu Chong et al. based on pre- and post-earthquake remote sensing image data through visual interpretation methods, and its accuracy was verified by field verification. The list contains 197,481 landslides and is included in the global earthquake-induced landslide database by the United States Geological Survey. Due to the lack of high-resolution images in some areas, there are certain errors in the identification of some landslides. Therefore, this embodiment selects a sub-area covered by high-resolution data as the research scope, which contains 196,008 landslides.

[0066] In this embodiment, 18 environmental factors affecting landslides were collected, including 5 factors related to the Wenchuan earthquake (peak ground acceleration, peak ground velocity, seismic intensity, distance from the epicenter of the earthquake, distance from the fault), 7 topographic factors (altitude, ground roughness, terrain relief, slope, aspect, topography, land cover), 4 geological and hydrological factors (lithology, vegetation cover index, average annual precipitation, distance from the nearest river) and 2 human engineering activity factors (distance from the nearest settlement, distance from the nearest road). Data sources include multiple authoritative institutions and public data sets, such as the United States Geological Survey (USGS), China National Earthquake Science Data Center (CENC), Chinese Academy of Sciences Resource and Environmental Science Data Center (RESDC), etc. Digital elevation model (DEM) data is used to extract terrain-related factors, and data such as lithology, rivers and roads are from the World Soil Data Center (ISRIC) and China National Basic Geographic Information Center.

[0067] In order to more accurately reflect the geomorphic characteristics of the study area, this embodiment uses an improved curvature basin method to divide the study area into 17,768 slope units. The landslide inventory data and related environmental factors are integrated into these slope units to provide input data for subsequent modeling. For the analysis of landslide intensity, the number of landslides in each unit is also counted, so as to achieve a joint evaluation of landslide susceptibility and intensity.

[0068] (2) Conduct collinearity test and calculate the contribution of each environmental factor to landslide according to the random forest model, and select the main controlling factors according to the contribution;

[0069] In this embodiment, the 18 collected environmental influencing factors are tested for collinearity, and at the same time, one of the variables with higher correlation is screened out in combination with the Spearman correlation coefficient to ensure that the VIF value is less than the threshold value of 10, and there is acceptable collinearity entering the model. After many experiments, the two factors of terrain undulation and peak ground velocity were deleted. Furthermore, in order to determine the most relevant influencing factors affecting landslides in the area and improve the accuracy and modeling efficiency of the model, the mainstream machine learning method of random forest is used to calculate the increase in node purity to obtain the importance ranking of factors, and the 16 factors that pass the collinearity test are further screened to obtain the factors that have the greatest impact on the landslide risk in the study area. The results are as follows: Figure 2 As shown. Figure 2 According to the results, this embodiment retains the relatively important first 11 influencing factors (the factors above the dotted line in the figure) as the main control factors to construct the subsequent hazard detection index system required, including roughness (ROU), distance to fault (DTF), altitude (ELE), distance to epicenter (DTE), distance to famous place (DTSE), precipitation (PRE), earthquake intensity (MI), normalized difference vegetation index (NDVI), slope (SLO), peak ground acceleration (PGA) and slope aspect (ASP).

[0070] (3) Construct a Bayesian spatial shared hyperparameter model (SSHM) considering landslide susceptibility and landslide intensity as shown in equations (1)-(3);

[0071] (4) using the Bayesian spatial shared hyperparameter model to output the comparable impact of shared environmental factors on disaster occurrence;

[0072] In this embodiment, the Bayesian spatial sharing hyperparameter model outputs the regression coefficient β of the influencing factor. k,q ,like Figure 3 As shown in , it is possible to obtain the regression coefficients of landslide susceptibility and landslide intensity at the same time. Figure 3It can be seen that the 11 environmental factors after screening have a significant impact on the landslide in the study area, and the credible interval is narrow, indicating that the uncertainty of the result is low. Further, the present embodiment finds that slope is the most important environmental factor affecting the susceptibility and intensity of landslides in the study area, and is positively correlated with landslide occurrence under a global scale. Except for slope, the four factors (MI, epicenter distance, fault, PGA) with the greatest impact are all earthquake-related factors, which are consistent with the data source of landslide disaster points. The relationship between MI and PGA, two factors reflecting the intensity of ground vibration, and the occurrence of landslides is positively correlated, indicating that the greater the intensity of ground vibration, the more likely landslides are to occur and the more times landslides occur. The relationship between epicenter distance and fault and the occurrence of landslides is negatively correlated, indicating that the area far away from the epicenter and the fault is relatively not prone to landslides. More specifically, thanks to the integrity of the independent modeling framework of the Bayesian spatial shared hyperparameter model described in the present invention, it can be found that the slope has a greater impact on the susceptibility of landslides and a relatively weaker impact on intensity; among the earthquake-related factors, the epicenter distance, faults, and PGA all have a relatively greater impact on the intensity of landslides, while MI has a greater impact on the susceptibility of landslides. In addition, precipitation and NDVI show a positive correlation with the occurrence of landslides in this study area, while ground roughness, elevation, distance from residential areas, and slope aspect show a negative correlation. Among them, NDVI and elevation have a greater impact on the susceptibility of landslides, and a very weak impact on intensity; while the four factors of ground roughness, distance from residential areas, precipitation, and slope aspect have almost the same impact on the susceptibility and intensity of landslides in this study area.

[0073] (5) identifying shared spatial random effects using the Bayesian spatial shared hyperparameter model;

[0074] The shared spatial random effects ξ of landslide susceptibility and intensity are estimated simultaneously using the Bayesian spatial shared hyperparameter model. i,q , the results are as follows Figure 4 As shown in Figure 2, the spatial autocorrelation trend in the study area is shown. Figure 4 It can be seen that the high-value areas are mainly concentrated near the fault zone in the northeast-southwest direction. These areas are greatly affected by earthquakes and show strong spatial dependence; while the low-value areas are located in the edge of the study area far away from the fault zone, where the randomness of landslides is high and the spatial effect is weak. Compared with susceptibility, the spatial random effect of intensity shows a smaller range of high-value areas, which may be related to the characteristic that earthquake intensity decays with distance. The mountainous and hilly terrain in the central part of the study area is more concentrated due to the impact of earthquakes. The random effect of intensity shows strong spatial aggregation in these areas, while the random effect in the marginal areas gradually weakens.

[0075] It is worth noting that the Bayesian spatial shared hyperparameter model of the present invention links the spatial effects of the two target variables (susceptibility and intensity) by sharing the hyperparameter distribution, so that each target variable retains an independent random effect distribution on the basis of shared information. This mechanism not only improves the flexibility of the model in capturing the spatial characteristics of landslides, but also significantly reduces the risk of overfitting, thereby generating a smoother and more regular spatial trend map. This trend map is more in line with the distribution characteristics of landslides driven by geology, earthquakes and geomorphology, and can provide a more scientific basis for regional landslide prevention and control planning.

[0076] In addition, the shared spatial random effect is only a part of the landslide hazard assessment, and the final susceptibility and intensity prediction results are the result of the shared random effect and the fixed effect of environmental factors. The Bayesian spatial shared hyperparameter model design of the present invention emphasizes the key role of environmental factors, such as the dominance of earthquake intensity, slope and precipitation in landslide hazard, while effectively avoiding the limitations brought by relying solely on spatial random effects.

[0077] (6) predicting landslide susceptibility and landslide intensity using the Bayesian spatial shared hyperparameter model;

[0078] The landslide susceptibility results obtained using the Bayesian spatial sharing hyperparameter model are as follows: Figure 5 As shown, the landslide susceptibility is classified into five categories (very high, high, medium, low, and very low) using the geometric breakpoint method, which is specially designed to handle continuous data, ensuring that the range of each class is roughly the same as the number of values ​​each class has, and the changes between intervals are very consistent. Figure 5 It can be seen that most areas show extremely high susceptibility to landslides, which is related to the huge impact of the Wenchuan earthquake. These extremely high susceptibility areas are concentrated in the central part of the study area, extending along the Longmenshan fault zone through the epicenter to Qingchuan County. The extremely low susceptibility areas of landslides are mainly concentrated in the plains in the southeast of the study area near Chengdu-Deyang-Mianyang, and there are also a small number of them in the mountainous areas in the southwest.

[0079] Using the same classification method, the results of landslide intensity and relative intensity obtained using the Bayesian spatial sharing hyperparameter model are as follows: Figure 6 In this embodiment, Where S i represents the area of ​​the i-th research unit (i.e., the slope unit in this embodiment), S I represents the total area of ​​the study area, and N represents the total number of landslides in the study area. Figure 6 It can be clearly found that compared Figure 5 Based on the susceptibility assessment results shown in the figure, landslide intensity hazard assessment is performed ( Figure 6(a)) After the extremely high-risk area is greatly reduced, the key intervention area where multiple landslides occur can be located under limited resource conditions. It can be further found that the areas with extremely high landslide intensity are concentrated in the mountainous area between the Maowen Fault Zone and the Longmenshan Fault Zone, and landslide units with smaller range and higher hazard (multiple landslides may occur) are identified, so that more accurate hazard location can be performed. The intensity hazard decreases successively from the extremely high-intensity area around the epicenter and the main fault zone to the edge of the study area.

[0080] Furthermore, this embodiment considers the influence of the area of ​​a single slope unit on the basis of the landslide intensity and fits the result of the relative intensity of the landslide ( Figure 6 (b)) shows some differences from the intensity results. It can be intuitively found that the relative intensity results ( Figure 6 (b) Comparison of medium-high hazard and extremely high hazard areas with traditional intensity results ( Figure 6 (a)) becomes less, which means that the areas that need intervention first can be more accurately located after the danger is triggered. In other words, the relative intensity is more focused on the real high-risk areas. Figure 6 From the small figure, we can find that the relative intensity not only identifies the higher hazard of the original extremely high intensity hazard area (multiple landslides and the number of occurrences is much higher than the expected value of landslides in this slope unit), but also newly identifies some small-area slope units with high relative intensity hazard. Although these areas are small in area and the total number of landslides is not large, the relative number is large, that is, the intensity hazard is not high but the relative intensity hazard is extremely high. Therefore, the traditional intensity hazard mapping results that do not consider the area of ​​the slope unit will ignore these areas with small slope units, few total landslides, but high relative hazard. In fact, in the actual disaster risk response, such places with high relative hazard should be paid more attention. In other words, the intensity results obtain the areas with a large total number of landslides at a macro level, while the relative intensity results more comprehensively identify the real high-risk areas where the number of landslides is higher than the expected value.

[0081] In the above embodiment, thanks to the independent unified framework of the Bayesian spatial shared hyperparameter model described in the present invention, the direct comparable impact of multiple environmental factors on the multidimensional danger of geological disasters can be obtained simultaneously: through the independent unified framework proposed by the present invention, the direct comparable analysis of multiple environmental factors on the susceptibility and intensity of geological disasters can be achieved. The 11 key environmental factors selected (such as slope, earthquake intensity, epicenter distance, etc.) show different contributions in the two hazard dimensions. For example, the impact of slope on susceptibility is significantly stronger than the impact on intensity, while earthquake intensity and peak ground acceleration (PGA) contribute more to intensity. This direct comparable analysis overcomes the limitation that the factor effects of traditional separation modeling methods cannot be compared, reveals the specific driving mechanism of multidimensional dangers of landslides, and provides a scientific basis for geological disaster management.

[0082] The present invention can fit the spatial shared random effects of multi-dimensional hazards of geological disasters: the present invention fits the shared spatial random effects of geological disaster susceptibility and intensity through the Bayesian spatial shared hyperparameter model, and clearly depicts the spatial distribution law of disasters in the region. The above-mentioned embodiment study found that high-value spatial effects are concentrated along the fault zone, while the marginal areas far away from the fault zone show low-value effects. The SSHM model of the present invention captures the correlation between target variables and their respective independent characteristics by sharing the hyperparameter distribution, significantly improving the model's ability to analyze spatial autocorrelation. Compared with the spatial shared component model (SSCM) that shares the entire random effect and its hyperparameters, the SSHM model emphasizes the role of environmental factors while retaining spatial characteristics, such as the contribution of key factors such as earthquake intensity, slope and precipitation. This improvement effectively reduces the overfitting problem caused by the excessive dominance of random effects, and provides a more scientific and reliable means for the study of the spatial distribution characteristics of disasters.

[0083] The present invention considers its inherent correlation through an independent and unified framework and obtains multi-dimensional hazard mapping at the same time (landslide susceptibility, landslide intensity, and landslide relative intensity in the above embodiment): The present invention uses the SSHM model to simultaneously generate a multi-dimensional hazard map of geological disasters. In the above embodiment, the susceptibility mapping clearly shows the high-risk areas along the fault zone in the middle of the study area, while the low-risk areas are mainly distributed in the marginal plain area. The spatial mapping of intensity and relative intensity further identifies the areas where multiple landslides have occurred and the small-area high-risk areas. As an innovative indicator, relative intensity can more accurately characterize the distribution of landslide intensity relative to the expected value of the unit. Compared with traditional intensity mapping, the relative intensity map identifies some landslide units with smaller areas but higher hazard levels. Although these areas have a small total number of landslides, they have become real high-risk areas due to their high relative frequency of occurrence. The multi-dimensional hazard mapping method of the present invention can more scientifically support disaster emergency and planning decisions under limited resources, and significantly improve the pertinence and efficiency of geological disaster management.

[0084] It should be noted that, in addition to the landslide-type geological disasters in the above-mentioned embodiments, other geological disasters such as mudslides can also be subjected to regional geological disaster multi-indicator hazard assessment through the regional geological disaster multi-indicator spatial hazard assessment model and its implementation method and application described in the present invention.

[0085] In summary, the present invention has broad application prospects, including:

[0086] (1) Regional disaster prevention and control planning

[0087] The present invention can be widely used in disaster prevention and control planning in mountainous and hilly areas. The generated disaster hazard map can accurately identify high-risk areas, guide local governments to optimize land use planning, formulate disaster prevention and mitigation policies, and provide clear intervention priorities. The present invention can be used as an important tool for hazard assessment in mountainous urban construction, highway route selection, reservoir and other large-scale infrastructure construction projects.

[0088] (2) Post-disaster recovery and emergency management

[0089] After a major disaster event, the present invention can quickly generate a disaster risk distribution map to provide decision support for post-disaster emergency response and resource allocation. For example, by identifying key areas where multiple landslides have occurred, it can help rescue teams prioritize deployment of forces and reduce casualties and economic losses. In addition, the relative intensity index can help managers accurately locate high-risk areas that require long-term monitoring and improve the scientific nature of post-disaster recovery.

[0090] (3) Ecological environmental protection and sustainable development

[0091] By accurately identifying high-risk areas for disasters, the present invention can effectively reduce the damage of disasters to the natural environment and reduce the possibility of secondary disasters caused by human activities. In addition, the generated spatial hazard map can be used as a reference for the design of ecological restoration projects, helping to optimize vegetation restoration plans and soil erosion control strategies, and helping to achieve the sustainable development goals of the region.

[0092] (4) Technical basis for multi-hazard joint assessment

[0093] Although the example of the present invention is aimed at a single geological disaster, the SSHM modeling framework proposed by it has strong versatility and can be extended to the disaster chain risk assessment of other geological disasters (such as debris flow, collapse, etc.). In areas where multiple disasters overlap, the method of the present invention can provide theoretical and technical support for joint disaster risk research and bring new solutions to risk management in a complex disaster context.

[0094] The present invention also has the social benefits of improving regional disaster response capabilities, reducing casualties and economic losses, and promoting the realization of regional comprehensive disaster reduction goals.

[0095] The independent unified modeling framework constructed by the present invention can integrate multi-source data, quickly generate multiple types of geological disaster hazard maps, accurately identify high-risk areas, and significantly improve the accuracy of disaster warning and prevention, thereby reducing the threat of landslide disasters to human society, providing local governments and emergency management departments with clear disaster spatial distribution information, and improving emergency response efficiency and capabilities. By promoting this technology, it can help disaster-prone areas establish a systematic disaster risk assessment system, reduce the long-term negative impact of disasters on economic and social development, and help achieve regional comprehensive disaster reduction and green sustainable development goals.

[0096] The above description is only a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any technician familiar with the profession can make some changes or modifications to equivalent embodiments of equivalent changes using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of the technical solution of the present invention.

Claims

1. A method for implementing a multi-index spatial risk assessment model for regional geological hazards, characterized in that: The following steps are involved: S1: Collect and preprocess geological disaster occurrence data and environmental impact factor data; S2: Calculate the contribution of each environmental impact factor to the occurrence of geological disasters, and screen out the main controlling factors according to the contribution; S3: Taking the geological disaster occurrence data as the dependent variable and the main controlling factors as the independent variables, a Bayesian spatial shared hyperparameter model is constructed that takes into account the susceptibility and intensity of geological disasters.

2. The method for implementing the regional geological disaster multi-index spatial risk assessment model according to claim 1 is characterized in that: In step S1, the environmental impact factors in the environmental impact factor data include earthquake factors, topographic factors, geological and hydrological factors, and human engineering activity factors.

3. The method for implementing the regional geological disaster multi-index spatial risk assessment model according to claim 2 is characterized in that: The earthquake triggering factors include peak ground acceleration, peak ground velocity, earthquake intensity, distance from the epicenter of the earthquake, and distance from the fault; The topographic factors include altitude, ground roughness, terrain relief, slope, aspect, topography and land cover; The geohydrological factors include lithology, vegetation cover index, average annual precipitation, and distance to the nearest river; The human engineering activity factors include the distance to the nearest settlement and the distance to the nearest road.

4. The method for implementing the regional geological disaster multi-index spatial risk assessment model according to claim 1 is characterized in that: In step S2, the random forest model is used to calculate the contribution of each environmental influencing factor to the occurrence of geological disasters.

5. The method for implementing the regional geological disaster multi-index spatial risk assessment model according to claim 1 is characterized in that: In step S2, before calculating the contribution, a collinearity test is also included to ensure that the VIF value is less than a threshold.

6. The method for implementing the regional geological disaster multi-index spatial risk assessment model according to claim 1 is characterized in that: In step S3, the Bayesian spatial sharing hyperparameter model is: Where: η i,q is the linear predictor of the target variable for each slope unit i in group q; g(·) is the linear predictor of η i,q Likelihood function connected with observed data; g q is the likelihood function of the q group of target variables; Y i (1) Y is the susceptibility to geological disasters; i (2) is the intensity of geological disasters; f(·) is a different potential Gaussian model; α q is the independent intercept term within group q; K is the number of influencing factors; β k,q is the grouped regression coefficient of the kth environmental covariate in group q; X k is the value of the kth influencing factor; ξ i,q is the shared spatial intercept random effect of the target variable of q groups; ε i,q is the residual of each group; i is the spatial intercept random effect for each spatial unit; ξ -i is the adjacent spatial unit of slope unit i; N(·) is a normal distribution; is the weight w in a given space i Under the condition of -i Mean; w i is the spatial adjacency matrix of spatial unit i; δ ξ is the variance of the spatial intercept random effect; is the total number of adjacent spatial units of spatial unit i; ξ j is the random effect value of the spatial intercept of the jth adjacent spatial unit of spatial unit i.

7. The method for implementing the regional geological disaster multi-index spatial risk assessment model according to claim 6 is characterized in that: The susceptibility of geological disasters obeys binomial distribution, and the intensity of geological disasters obeys Poisson distribution.

8. A multi-index spatial risk assessment model for regional geological hazards, characterized in that: It is established by adopting the implementation method of the regional geological disaster multi-index spatial risk assessment model described in any one of claims 1-7.

9. Application of the multi-index spatial risk assessment model for regional geological disasters as described in claim 8 in multi-index risk assessment for regional geological disasters.

10. The application of the regional geological disaster multi-index spatial risk assessment model according to claim 9 in the regional geological disaster multi-index risk assessment is characterized in that: The assessment of the multi-index risk of regional geological hazards specifically includes the following steps: Using the Bayesian spatial shared hyperparameter model to output the comparable impacts of shared environmental factors on disaster occurrence; identifying shared spatial random effects using the Bayesian spatial shared hyperparameter model; The Bayesian spatial sharing hyperparameter model is used to predict the susceptibility and intensity of geological disasters.

Images