Prediction and forecasting methods suitable for mountain tsunamis with various generation mechanisms

Abstract

This provides prediction and forecasting methods suitable for mountain tsunamis with various generation mechanisms. [Solution] The method takes into account the differences in the generation mechanisms of landslides and proposes an integrated prediction and forecasting model for multi-type landslides with physical mechanisms, an optimization strategy for its parameters, and evaluation indicators. This extends the conventional prediction and forecasting of single-type landslides to prediction and forecasting of multi-type landslides, providing the main controlling factors and prediction and forecasting processes for the generation, development, and damage of landslides in any given region. This allows for accurate prediction of the physical processes of generation and development of different types of landslides, is highly applicable, and can better support decision-making for landslide disaster countermeasures.

Description

[Technical Field]

[0001] This invention belongs to the technical fields of hydrology and meteorological forecasting, and more particularly to a prediction and forecasting method suitable for landslides with various generation mechanisms. [Background technology]

[0002] Mountain tsunamis are the most common type of flood in Japan, causing the most casualties and economic losses. They are characterized by their numerous locations, wide distribution, suddenness, high death toll, and significant regional differences. As abnormal rainfall events increase, the risk of mountain tsunamis rises, making accurate prediction and forecasting of mountain tsunamis a crucial technical tool for improving tsunami mitigation capabilities. However, mountain tsunamis are not only natural phenomena but also social phenomena. Their occurrence, development, and damage are the result of a complex interplay of many factors, including rainfall, supporting structure surfaces, and human activities. These factors exhibit high spatiotemporal heterogeneity, resulting in complex mechanisms of mountain tsunami occurrence and development in small watersheds across different regions. Furthermore, different types of mountain tsunamis have different causes, potential for damage, and destructiveness. Ignoring these differences in mountain tsunami occurrence mechanisms leads to uncertainty in mountain tsunami prediction.

[0003] Currently, simulations, predictions, and forecasts of landslides in small watersheds primarily employ a single hydrological or fluid dynamics model, and focus mainly on the processes of a single type of landslide, such as those frequently triggered by short-duration heavy rainfall. However, recent landslide disaster events reveal that chain reactions and complex interactions of processes such as rapid water level rise due to flooding in small watersheds, sediment erosion and deposition, and glacial melting are particularly prominent, and that events where small amounts of rainfall trigger large-scale disasters frequently occur. Therefore, prediction and forecasting methods for a single type of landslide do not take into account the differences in landslide generation mechanisms, making it difficult to accurately reproduce the generation and development processes of other different types of landslides. This results in low accuracy in landslide prediction and forecasting, significantly impacting decision-making in landslide disaster countermeasures. Consequently, it is urgent to develop prediction and forecasting methods suitable for landslides with various generation mechanisms, improve the accuracy of landslide prediction, and reduce the risks and losses associated with landslides. [Overview of the project] [Problems that the invention aims to solve]

[0004] The objective of this invention is to provide a prediction and forecasting method suitable for mountain tsunamis with various generation mechanisms, thereby solving the above-mentioned problems. [Means for solving the problem]

[0005] To achieve the above objectives, the present invention provides the following technical solutions. The present invention is a prediction and forecasting method suitable for mountain tsunamis with various generation mechanisms, The steps involve collecting and organizing basic data for the target area, including various types of basic data for the target area, such as precipitation from multiple precipitation sources, water level processes, flow rate processes, sediment load processes, snowmelt processes, weather, distribution of villages along rivers, gully measurements, basic geographic information, and data from past landslide disaster damage surveys. The precipitation data from multiple precipitation sources includes precipitation observed at observation stations, precipitation quantitatively estimated by radar, and precipitation data retrieved from satellite data. The data from past landslide disaster damage surveys includes the number of deaths / missing persons based on the damage, the number of collapsed houses, the amount of direct economic damage, flood trace data, and flooded area. The flood trace data includes the location of flood traces and the time of occurrence of flood traces. Step 1 involves extracting maximum values ​​from the water level process, flow rate process, and sediment load process for each event, including the elevation of flood traces, determining the peak water level, peak flow rate, and peak sediment load values ​​for each event, calculating the sum of the snowmelt processes for each event, determining the total snowmelt amount for each event, then dividing the area based on basic geographic information data and distribution data of villages along the river, and based on a set area threshold, to obtain basic calculation units for partial watersheds, extracting the distribution of the river system and its water collection topology, dividing the obtained partial watershed basic calculation units based on a set water collection area threshold and distribution data of villages along the river, to obtain watershed units, and establishing the water collection topology for each watershed unit. Step 2 involves determining the type classification of mountain tsunamis and their respective catchment topologies, first analyzing the main control factors in the process of occurrence, development, and damage occurrence of mountain tsunamis in each watershed based on the peak water level of floods, peak flood flow rates, peak sediment load values, total snowmelt, and damage survey data of past mountain tsunami disasters for each event in the target area; secondly, determining the type classification of mountain tsunamis in the target area by employing a clustering algorithm based on the main control factors and their values ​​for each watershed unit, the mountain tsunami type classification of the target area including heavy rain type mountain tsunami classification, snowmelt type mountain tsunami classification, water flow / sediment type mountain tsunami classification, rain-snow composite type mountain tsunami classification, and snow-sediment composite type mountain tsunami classification; and further establishing the catchment topology for each mountain tsunami type classification based on the catchment topology of each watershed unit and the determined mountain tsunami type classification. The steps involve constructing an integrated prediction and forecasting model for multi-type landslides, setting the model type step in hours or minutes, the model's spatial scale being the basic calculation unit for a partial watershed, establishing a short-term numerical precipitation forecasting model, a nonlinear runoff generation and confluence model, a snowmelt runoff generation and confluence model, and one-dimensional and two-dimensional fluid dynamics models for water flow and sediment of heterogeneous landslides, determining the input and output data for each model, the input data for the short-term numerical precipitation forecasting model being precipitation quantitatively estimated by radar and precipitation data retrieved from satellite data, and the output data being The short-term forecast precipitation data is used, the input data for the nonlinear runoff generation / confluence model is precipitation and meteorological data observed at observation stations, and the output data is gully flow rate process data; the input data for the snowmelt runoff generation / confluence model is precipitation and meteorological data observed at observation stations, and the output data is gully flow rate and snowmelt process data; the input data for the one-dimensional and two-dimensional fluid dynamics models of water flow and sediment in heterogeneous flow tsunamis is precipitation and gully measurement data observed at observation stations, and the output data is gully flow rate, water level, sediment load, and inundation depth process data. Step 3 involves performing normalization processing of input / output interfaces and model structures for each established model to construct a library of all types of landslide models, selecting a corresponding computational model from the library of all types of landslide models based on the determined landslide type classification, combining the computational models according to the order of precipitation-runoff occurrence / confluence-flood development in a partial watershed to establish a landslide prediction / forecasting model for each landslide type classification, extracting maximum values ​​from the gully water level process, flow rate process, sediment load process, and inundation depth process, respectively, using the output data of the landslide prediction / forecasting model for each classification, determining the predicted peak flood water level, peak flood flow rate, peak sediment load value, and maximum inundation depth for each classification, calculating the sum of the gully snowmelt processes to determine the predicted total snowmelt for each classification, determining the upstream and downstream type classification models of the landslide prediction / forecasting model for each classification based on the established catchment topology of each landslide type classification, and constructing a multi-type landslide prediction / forecasting integrated model for the target area. Steps for determining the optimal parameter set and evaluation index of the integrated model, in which, based on the catchment topology of each established type classification of tsunami, the tsunami prediction and forecasting models of each type classification of tsunami are calibrated one by one in the order from upstream to downstream to determine the optimal parameter set of the multi-type tsunami prediction and forecasting integrated model. Adopting a weighted comprehensive index to evaluate the accuracy of the multi-type tsunami prediction and forecasting integrated model and the tsunami prediction and forecasting models of each classification, the optimal value is 1, and the calculation formula is

Number

Number

Number

Number

Number

[0006] Furthermore, the water level, flow rate, sediment load, and snowmelt process data described in Step 1 are all data observed at observation stations; the meteorological data includes temperature, soil moisture content, surface evaporation, sunshine duration, and wind speed data observed at observation stations; the gully measurement data includes cross-sectional, longitudinal section, and bridge / culvert measurement data; the basic geographic information data includes a digital elevation model (DEM) with a scale of 1:50000 or higher, land use classification vector data, and soil property classification vector data; and the set area threshold is 50 km². 2 The set catchment area threshold is 200 km². 2 That is the case.

[0007] Furthermore, the specific process for establishing the watershed topology for each watershed unit described in Step 1 is to designate the watershed unit where the river source is located as the source basin, determine the river into which the river flows at the outlet of the source basin based on the water system's watershed topology, and repeat the process until there are no more rivers flowing into the outlet of the watershed unit, with the watershed unit where that river is located being the downstream watershed of the source basin.

[0008] Furthermore, the specific process of analyzing the main controlling factors in the occurrence, development, and damage processes of landslides in each watershed as described in Step 2 involves employing data mining using multivariate analysis to determine the contribution rate of each single factor—peak flood water level, peak flood flow rate, peak sediment load, and total snowmelt—to the data of the number of deaths / missing persons, number of collapsed houses, direct economic damage, elevation of the maximum flood trace, and maximum inundation area damage for each event in each watershed unit, as well as the corresponding damage for various watershed combination factors of landslides in each event in each watershed unit. The contribution rate to the situational data was analyzed, including combinations of peak flood flow rate and total snowmelt, combinations of peak flood flow rate, peak flood water level and peak sediment load, and combinations of total snowmelt, peak flood water level and peak sediment load. When the contribution rate of a single factor in the watershed exceeds 0.5, it is considered that the single factor in the watershed is the main controlling factor in the process of occurrence, development, and damage of landslides in each watershed. When the contribution rate of combined factors in the watershed is greater than the sum of the contribution rates of the single factors, it is considered that the combined factors in the watershed are the main controlling factors of landslides in the watershed. The formula for calculating the contribution rate is:

number

[0009] Furthermore, the specific process for establishing the water collection topology for each type of landslide described in Step 2 involves starting from the source area, analyzing in the order of water collection from the upstream to the downstream basin, and determining the type of landslide designated as Category A in one basin unit and the type of landslide designated as Category B in a downstream basin unit if they do not match. In such cases, the downstream type of Category A is determined to be Category B, and the upstream type of Category B is determined to be Category A. If the two match, Category A and Category B are merged into one hypothetical category. The search continues in downstream basin units until a downstream basin unit that does not match this landslide type category is found, visiting all basin units until the basin unit at the outlet of the target area is reached, and finally determining the upstream and downstream type of each type of landslide.

[0010] Furthermore, the specific process for normalizing the input / output interface and model structure for each established model described in Step 3 involves normalizing station-scale input data in the format of station code, station longitude, station latitude, type step, precipitation observed at the station, weather, snowmelt, and sediment load; normalizing precipitation quantitatively estimated by radar and precipitation data retrieved from satellite data according to the fourth type of grid data format of MICAPS; and normalizing gully measurement data in the format of cross-section code, longitude, latitude, and elevation. The process involves performing normalization on the output short-term forecast precipitation data according to MICAPS's fourth type of grid data format, normalizing the output gully flow process data in the format of model time step, current time step, model code, gully code, flow rate, water level, snowmelt amount, sediment load, and inundation depth data, normalizing the model file with basic information, parameter information, and model information, and adopting a unified input / output interface to generate a dynamic integration framework for various types of models, and performing model registration / deployment, encapsulation / release, and retrieval.

[0011] Furthermore, the specific process for constructing an integrated prediction and forecasting model for multi-type landslides in the target area described in Step 3 is as follows: For landslides with multiple upstream type classifications, the inundation depth process of all upstream watershed classifications is used as the upper boundary of the landslides with respect to

[0012] Furthermore, the specific process for determining the optimal parameter set for the multi-type landslide prediction and forecasting integrated model described in Step 4 involves optimizing the parameters of the landslide prediction and forecasting model for the heavy rain type landslide by adopting the flow rate process data collected in Step 1, optimizing the parameters of the landslide prediction and forecasting model for the snowmelt type and rain-snow combined type landslide by adopting the flow rate and snowmelt process data, and optimizing the parameters of the landslide prediction and forecasting model for the water flow-sediment type landslide by adopting the flow rate, water level and sediment load process data and flood trace data. The process involves optimizing the parameters of the tsunami prediction and forecasting model for snow-sedipal composite tsunami classifications by employing flow rate, water level, snowmelt amount, and sediment load process data, as well as flood trace data. The optimal parameters for each corresponding classification model are determined, and based on the determined optimal parameters for each classification model, a regional analysis method for parameters is employed to determine the parameters for tsunami classification models where the data observed at observation stations and flood trace data are insufficient. Finally, the optimal parameter set for the integrated model is determined using the tsunami type classification as a unit.

[0013] Furthermore, the method for localization analysis of the parameters is a spatial similarity method, an attribute similarity method, a parameter regression method, or a classification tree / regression tree method. [Effects of the Invention]

[0014] The beneficial effects of the present invention are as follows: The method of the present invention takes into account the differences in the generation mechanisms of landslides, employs a method that combines data mining by multivariate analysis with multiple types of numerical simulations, proposes an integrated prediction and forecasting model for multi-type landslides with physical mechanisms, an optimization strategy for its parameters, and evaluation indicators, extending conventional prediction and forecasting of single-type landslides to prediction and forecasting of multi-type landslides, providing the main controlling factors and prediction and forecasting processes for the generation, development, and damage occurrence of landslides in any given region, enabling accurate prediction of the physical processes of generation and development of different types of landslides, allowing for more comprehensive analysis, high applicability, and better supporting decision-making for landslide disaster countermeasures.

[0015] The present invention will be described in more detail below with reference to the drawings and specific embodiments. [Brief explanation of the drawing]

[0016] [Figure 1] This is a flowchart of the method described in the present invention. [Figure 2] This is a schematic diagram illustrating the process for determining the type classification of landslides and their respective water collection topologies. [Figure 3] This is a schematic diagram illustrating the process of constructing an integrated prediction and forecasting model for multi-type landslides. [Modes for carrying out the invention]

[0017] The present invention discloses a prediction and forecasting method suitable for mountain tsunamis with various generation mechanisms, and as shown in Figures 1 to 3, the method includes the following steps 1 to 5. Step 1 involves collecting and organizing basic data for the target area.

[0018] The study primarily collects and organizes various types of basic data for the target area, including data on precipitation from multiple sources, water level processes, flow processes, sediment load processes, snowmelt processes, weather, distribution of villages along rivers, gully measurements, basic geographic information, and damage surveys from past landslide disasters. Maximum values ​​are extracted from the water level processes, flow processes, and sediment load processes for each event, determining the peak water level, peak flow rate, and peak sediment load values ​​for each event. The sum of the snowmelt processes for each event is calculated to determine the total snowmelt amount for each event. Here, precipitation data from multiple sources includes precipitation observed at observation stations, precipitation quantitatively estimated by radar, and precipitation data retrieved from satellite data. The data on water content, water level, flow rate, sediment load, and snowmelt process are all data observed at observation stations; the meteorological data includes temperature, soil moisture content, surface evaporation, sunshine duration, and wind speed observed at observation stations; the gully measurement data includes cross-sectional, longitudinal section, and bridge / culvert measurement data; the basic geographic information data includes digital elevation models (DEM) with a scale of 1:50000 or higher, land use classification vector data, and soil property classification vector data; the data on damage from past landslides and tsunamis includes the number of dead / missing persons, number of collapsed houses, direct economic damage, flood trace data, and flooded area based on the damage situation; and the flood trace data includes the location of the flood trace, the time of occurrence of the flood trace, and the elevation of the flood trace.

[0019] Next, based on basic geographical information data of the target area and distribution data of villages along the river, 50km 2 The area is divided using an area threshold to obtain basic calculation units for the partial watershed, extract the distribution of the water system and its water collection topology, and then, using the basic calculation units for the divided partial watershed, 200 km 2 Based on the catchment area threshold and the distribution data of villages along the river, the system is divided to obtain watershed units, and the catchment topology of each watershed unit is established. Specifically, the watershed unit where the river source is located is designated as the source area, and based on the water system's catchment topology, the river into which the river flows at the outlet of the source area is determined. The watershed unit where this river is located is the downstream watershed of the source area, and this process is repeated until there are no more rivers flowing into the outlet of the watershed unit.

[0020] Step 2 involves determining the type of landslide and its corresponding water collection topology.

[0021] First, based on peak flood water levels, peak flood flow rates, peak sediment load values, total snowmelt, and damage survey data from past landslide disasters in the target area, we will analyze the main controlling factors in the process of landslide occurrence, development, and damage in each river basin. Specifically, we will employ data mining using multivariate analysis such as geodetector and sorting analysis to determine the contribution rate of each single factor—peak flood water level, peak flood flow rate, peak sediment load value, and total snowmelt—to corresponding damage data such as the number of deaths / missing persons, number of collapsed houses, direct economic damage, elevation of the maximum flood trace, and maximum inundation area for each landslide disaster event in each river basin unit, and to the corresponding damage data of various river basin combination factors for each landslide disaster event in each river basin unit. The contribution rates are analyzed, including combinations of peak flood flow rate and total snowmelt, combinations of peak flood flow rate, peak flood water level and peak sediment load, and combinations of total snowmelt, peak flood water level and peak sediment load. If the contribution rate of a single factor in the basin exceeds 0.5, then the single factor in the basin is the main controlling factor in the process of occurrence, development, and damage of landslides in each basin. If the contribution rate of combined factors in the basin is greater than the sum of the contribution rates of the single factors, then the combined factors in the basin are the main controlling factors of landslides in the basin. The formula for calculating the contribution rate is:

number

[0022] Next, based on the main control factors and their values ​​for each watershed unit, a clustering algorithm is employed to determine the main types and classifications of landslides in the target area. The main types of landslides include heavy rainfall type, snowmelt type, water flow / sediment type, rain-snow combined type, and snow-sediment combined type. As a result, the corresponding landslide type classifications are heavy rainfall type landslide classification, snowmelt type landslide classification, water flow / sediment type landslide classification, rain-snow combined type landslide classification, and snow-sediment combined type landslide classification. Furthermore, based on the water collection topology of each watershed unit and the determined landslide type classifications, the water collection topology for each landslide type classification is established, specifically starting from the source area and progressing from the upstream watershed to the downstream watershed. The analysis is performed in the order in which water is collected. If the type classification of a landslide in one watershed unit (labeled as Classification A) does not match the type classification of a landslide in a downstream watershed unit (labeled as Classification B), the downstream type classification of Classification A is determined to be Classification B, and the upstream type classification of Classification B is determined to be Classification A. If the two do match, Classification A and Classification B are merged into one hypothetical classification. The search continues in downstream watershed units until a downstream watershed unit that does not match this hypothetical classification is found, and all watershed units are visited until the watershed unit at the outlet of the target area is reached, and finally the upstream and downstream type classifications of each landslide classification are determined.

[0023] Step 3 involves constructing an integrated model for predicting and forecasting multi-type landslides.

[0024] The model type step is set in hours or minutes, the model's spatial scale is the basic calculation unit of a partial watershed, and short-term numerical precipitation forecasting models, nonlinear runoff generation and confluence models, snowmelt runoff generation and confluence models, and one-dimensional and two-dimensional fluid dynamics models of water flow and sediment for heterogeneous runoff tsunamis are established. The input and output data for each model are determined, where the input data for the short-term numerical precipitation forecasting model is precipitation quantitatively estimated by radar and precipitation data retrieved from satellite data, and the output data is short-term forecast precipitation data. The input data for the nonlinear runoff generation and confluence model is precipitation and meteorological data observed at observation stations, and the output data is gully flow process data. The input data for the snowmelt runoff generation and confluence model is precipitation and meteorological data observed at observation stations, and the output data is gully flow and snowmelt process data. The input data for the one-dimensional and two-dimensional fluid dynamics models of water flow and sediment in heterogeneous flow tsunamis is precipitation and gully measurement data observed at observation stations, and the output data is gully flow, water level, sediment load, and inundation depth process data.

[0025] For each established model, the input / output interface and model structure are normalized. Specifically, station-scale input data is normalized in the form of station code, station longitude, station latitude, type step, precipitation observed at the station, weather, snowmelt, and sediment load. Precipitation quantitatively estimated by radar and precipitation data retrieved from satellite data are normalized according to the fourth type of grid data format of MICAPS. Gully measurement data is normalized in the form of cross-section code, longitude, latitude, and elevation. The output short-term forecast precipitation data is then normalized. The system then performs normalization processing according to the fourth type of grid data format of MICAPS, normalizing the output gully flow process data in data formats such as model time step, current time step, model code, gully code, flow rate, water level, snowmelt amount, sediment load, and inundation depth, normalizing the model file in terms of basic information, parameter information, and model information, adopting a unified input / output interface to generate a dynamic integrated framework for various types of models, registering, deploying, encapsulating, releasing, and calling models, and further constructing a library of all types of landslide models.

[0026] Based on the determined landslide type classification, a corresponding calculation model is selected from the entire landslide model library. The calculation models are combined according to the sequence of precipitation-runoff occurrence / confluence-flood development in the partial watershed to establish a landslide prediction / forecasting model for each landslide type classification, as shown in Table 1. Next, using the output data of the landslide prediction / forecasting model for each classification, the maximum values ​​are extracted from the gully water level process, flow rate process, sediment load process, and inundation depth process, respectively. The predicted peak flood water level, peak flood flow rate, peak sediment load value, and maximum inundation depth for each classification are determined, and the sum of the gully snowmelt processes is calculated to determine the predicted total snowmelt for each classification. [Table 1]

[0027] Based on the established catchment topology of each type of tsunami, the upstream and downstream type classification models for the tsunami prediction and forecasting model for each classification are determined, and an integrated prediction and forecasting model for multi-type tsunamis in the target area is constructed. The integrated strategy is as follows: For tsunami types with multiple upstream type classifications, the inundation depth process of all upstream watershed classifications is used as the upper boundary of the tsunami type classification, and the sum of the calculation results of the remaining models excluding the inundation depth of all upstream type classifications is used as the input value for the gully of the tsunami type classification. For tsunami types with multiple downstream type classifications, the inundation depth process of the tsunami type classification is used as the lower boundary of all downstream type classifications, and the output value of the remaining gully excluding the inundation depth of the tsunami type classification is determined according to the weight of the water-carrying cross-sectional area of ​​the gully of the downstream type classification.

[0028] In Step 4, the optimal parameter set and evaluation metrics for the integrated model are determined.

[0029] Based on the established water collection topology for each type of tsunami, the tsunami prediction and forecasting models for each type are calibrated one by one from upstream to downstream, and an optimization strategy for the parameters of the integrated multi-type tsunami prediction and forecasting model for the target area is generated. Specifically, the parameters of the tsunami prediction and forecasting model for the heavy rain type are optimized by adopting the flow process data observed at observation stations collected in Step 1, the parameters of the snowmelt type and rain-snow combined type tsunami prediction and forecasting models are optimized by adopting the flow and snowmelt process data observed at observation stations, and the parameters of the water flow-sediment type tsunami are optimized by adopting the flow, water level, and sediment load process data and flood trace data observed at observation stations. The parameters of the tsunami prediction and forecasting model for each tsunami classification are optimized. Using flow rate, water level, snowmelt amount, and sediment load process data observed at observation stations, along with flood trace data, the parameters of the tsunami prediction and forecasting model for snow-sediment composite tsunami classifications are optimized. The optimal parameters for the model corresponding to each classification are determined. Based on the determined optimal parameters for each classification model, parameter localization analysis methods such as spatial similarity, attribute similarity, parameter regression, or classification tree / regression tree methods are employed to determine the parameters of tsunami classification models where the data observed at observation stations and flood trace data are insufficient. The optimal parameter set for the integrated model is then determined, using the tsunami type classification as a unit.

[0030] The accuracy of the integrated model and the segmented model was evaluated using a weighted composite index, and the optimal value was 1. The calculation formula is as follows:

number

number

number

number

number

[0031] Step 5 involves forecasting the classification of each type of tsunami and the process of the tsunami at its exit point.

[0032] Based on precipitation data from multiple precipitation sources, meteorological data, and gully measurement data collected in Step 1, a multi-type landslide prediction and forecasting integrated model is operated to forecast the type classification of each landslide within the target area, as well as the short-term forecast precipitation at the exit point of the target area, gully flow rate, water level, sediment load, snowmelt amount, and inundation depth process.

[0033] It should be noted that the above description does not limit the technical solutions of the present invention, but merely explains them. While the present invention is described in detail with reference to preferred configurations, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced with equivalent means without departing from the spirit and scope of the present invention.

Claims

1. A prediction and forecasting method suitable for mountain tsunamis with various generation mechanisms, The steps involve collecting and organizing basic data for the target area, including various types of basic data for the target area, such as precipitation from multiple precipitation sources, water level processes, flow rate processes, sediment load processes, snowmelt processes, weather, distribution of villages along rivers, gully measurements, basic geographic information, and data from past landslide disaster damage surveys. The precipitation data from multiple precipitation sources includes precipitation observed at observation stations, precipitation quantitatively estimated by radar, and precipitation data retrieved from satellite data. The data from past landslide disaster damage surveys includes the number of deaths / missing persons based on the damage, the number of collapsed houses, the amount of direct economic damage, flood trace data, and flooded area. The flood trace data includes the location of flood traces and the time of occurrence of flood traces. Step 1 involves extracting maximum values ​​from the water level process, flow rate process, and sediment load process for each event, including the elevation of flood traces, determining the peak water level, peak flow rate, and peak sediment load values ​​for each event, calculating the sum of the snowmelt processes for each event, determining the total snowmelt amount for each event, then dividing the area based on basic geographic information data and distribution data of villages along the river, and based on a set area threshold, to obtain basic calculation units for partial watersheds, extracting the distribution of the river system and its catchment topology, dividing the obtained partial watershed units based on a set catchment area threshold and distribution data of villages along the river, to obtain watershed units, and establishing the catchment topology for each watershed unit. Step 2 involves determining the type classification of mountain tsunamis and their respective catchment topologies, first analyzing the main control factors in the process of occurrence, development, and damage occurrence of mountain tsunamis in each watershed based on the peak water level of floods, peak flood flow rates, peak sediment load values, total snowmelt, and damage survey data of past mountain tsunami disasters for each event in the target area; secondly, determining the type classification of mountain tsunamis in the target area by employing a clustering algorithm based on the main control factors and their values ​​for each watershed unit, the mountain tsunami type classification of the target area including heavy rain type mountain tsunami classification, snowmelt type mountain tsunami classification, water flow / sediment type mountain tsunami classification, rain-snow combined type mountain tsunami classification, and snow-sediment combined type mountain tsunami classification; and further, establishing the catchment topology for each mountain tsunami type classification based on the catchment topology of each watershed unit and the determined mountain tsunami type classification. The steps involve constructing an integrated prediction and forecasting model for multi-type landslides, setting the model type step in hours or minutes, the model's spatial scale being the basic calculation unit for a partial watershed, establishing a short-term numerical precipitation forecasting model, a nonlinear runoff generation and confluence model, a snowmelt runoff generation and confluence model, and one-dimensional and two-dimensional fluid dynamics models for water flow and sediment of heterogeneous landslides, determining the input and output data for each model, the input data for the short-term numerical precipitation forecasting model being precipitation quantitatively estimated by radar and precipitation data retrieved from satellite data, and the output data being The short-term forecast precipitation data is used, the input data for the nonlinear runoff generation / confluence model is precipitation and meteorological data observed at observation stations, and the output data is gully flow rate process data; the input data for the snowmelt runoff generation / confluence model is precipitation and meteorological data observed at observation stations, and the output data is gully flow rate and snowmelt process data; the input data for the one-dimensional and two-dimensional fluid dynamics models of water flow and sediment in heterogeneous flow tsunamis is precipitation and gully measurement data observed at observation stations, and the output data is gully flow rate, water level, sediment load, and inundation depth process data. Step 3 involves performing normalization processing of input / output interfaces and model structures for each established model to construct a library of all types of landslide models, selecting a corresponding computational model from the library of all types of landslide models based on the determined landslide type classification, combining the computational models according to the order of precipitation-runoff occurrence / confluence-flood development in a partial watershed to establish a landslide prediction / forecasting model for each landslide type classification, extracting maximum values ​​from the gully water level process, flow rate process, sediment load process, and inundation depth process using the output data of the landslide prediction / forecasting model for each classification, determining the predicted peak flood water level, peak flood flow rate, peak sediment load value, and maximum inundation depth for each classification, calculating the sum of the gully snowmelt processes to determine the predicted total snowmelt for each classification, determining the upstream and downstream type classification models of the landslide prediction / forecasting model for each classification based on the established catchment topology of each landslide type classification, and constructing a multi-type landslide prediction / forecasting integrated model for the target area. The step of determining the optimal parameter set and evaluation indicators for the integrated model is to calibrate each tsunami prediction / forecasting model for each tsunami type, starting from upstream to downstream, based on the established catchment topology for each tsunami type, and to determine the optimal parameter set for the multi-type tsunami prediction / forecasting integrated model. A weighted composite index was used to evaluate the accuracy of the integrated prediction and forecasting model for multi-type landslides and the landslide prediction and forecasting models for each category. The optimal value was 1, and the calculation formula is as follows: [Math 1] And, Here, g is a weighted composite index of a multi-type landslide forecasting and prediction integrated model, m is the number of landslide type classifications in the target area, i is the i-th landslide type classification for which data and flood trace data are available at observation stations in the target area, and 1 ≤ i ≤ m, g ｉ and α ｉ These are the weighted overall index and its weight for the i-th category, respectively. [Math 2] RMSE ｉ is the root mean square error of the precipitation in the i-th segment, [Math 3] P ｏ，ｊ，ｉ and P ｓ，ｊ，ｉ These represent the precipitation observed at the observation station at the j-th time of the i-th segment and the short-term forecast precipitation output by the model, respectively. ｉ is the sequence length of precipitation data observed at the i-th observation station, and |Re ｉ | represents the absolute value of the relative error in the peak flood flow rate, the relative error in the peak flood water level, the relative error in the peak sediment load, the relative error in the amount of snowmelt, or the relative error in the inundation depth for the i-th category. [Math 4] and Q ｏ，ｐ，ｉ is the peak discharge of flood or the peak water level of flood or the peak sediment load value or the total snowmelt volume or the elevation of flood mark determined by the data observed at the observation site in the i-th section, and Q ｓ，ｐ，ｉ is the peak discharge of flood or the peak water level of flood or the peak sediment load value or the total snowmelt volume or the maximum inundation depth predicted by the model in the i-th section, and NSE ｉ is the Nash efficiency coefficient of the flow hydrograph or the Nash efficiency coefficient of the water level hydrograph or the Nash efficiency coefficient of the sediment load hydrograph or the Nash efficiency coefficient of the snowmelt volume hydrograph in the i-th section, [Math 5] Q ｏ，ｊ，ｉ Q is the flow rate, water level, sediment load, or snowmelt amount observed at the observation station at the j-th time of the i-th division. ｓ，ｊ，ｉ This is the flow rate, water level, sediment load, or snowmelt amount predicted by the model for the j-th time in the i-th segment. 【number】 This is the average flow rate, average water level, average sediment load, or average snowmelt rate observed at the i-th section of the observation station, and N ｉ γ is the sequence length of the flow rate, water level, sediment load, or snowmelt amount observed at the i-th section observation station, and γ ｉ is the weight of the precipitation assessment index for the i-th segment, and β ｉ Step 4 is the weight of the flow rate, water level, sediment load, snowmelt amount, or inundation depth evaluation index for the i-th category, Step 5 is a step of forecasting the type classification of each landslide and the process of the landslide at its exit, wherein, based on precipitation data from multiple precipitation sources, meteorological data, and gully measurement data collected in Step 1, a multi-type landslide prediction and forecasting integrated model is operated to forecast the type classification of each landslide within the target area and the short-term forecast precipitation, gully flow rate, water level, sediment load, snowmelt amount, and inundation depth process at the exit of the target area. A method characterized by including the following.

2. The water level, flow rate, sediment load, and snowmelt process data described in Step 1 are all data observed at observation stations; the meteorological data includes temperature, soil moisture content, surface evaporation, sunshine duration, and wind speed data observed at observation stations; the gully measurement data includes cross-sectional, longitudinal section, and bridge / culvert measurement data; the basic geographic information data includes a digital elevation model (DEM) with a scale of 1:50000 or higher, land use classification vector data, and soil property classification vector data; and the set area threshold is 50 km². ２ The set catchment area threshold is 200 km. ２ A prediction and forecasting method suitable for mountain tsunamis with various generation mechanisms as described in feature 1.

3. The specific process for establishing the water collection topology of each watershed unit described in Step 1 is to designate the watershed unit where the river source is located as the source area, determine the river into which the river flows at the outlet of the source area based on the water system's water collection topology, and repeat the process until there are no more rivers flowing into the watershed unit where the river flows at the outlet of the watershed unit, as described in 1.

4. The specific process for analyzing the main controlling factors in the occurrence, development, and damage processes of landslides in each watershed as described in Step 2 involves employing data mining using multivariate analysis to determine the contribution rate of each single factor—peak flood water level, peak flood flow rate, peak sediment load, and total snowmelt—to the corresponding data of deaths / missing persons, number of collapsed houses, direct economic damage, elevation of the maximum flood trace, and maximum inundation area damage for each event in each watershed unit, as well as the corresponding damage status of various watershed combination factors for each event in each watershed unit. The contribution rates to the data were analyzed, including combinations of peak flood flow rate and total snowmelt, combinations of peak flood flow rate, peak flood water level and peak sediment load, and combinations of total snowmelt, peak flood water level and peak sediment load. When the contribution rate of a single factor in the watershed exceeds 0.5, it is considered that the single factor in the watershed is the main controlling factor in the process of occurrence, development, and damage of landslides in each watershed. When the contribution rate of combined factors in the watershed is greater than the sum of the contribution rates of the single factors, it is considered that the combined factors in the watershed are the main controlling factors of landslides in the watershed. The formula for calculating the contribution rate is: [Math 6] And, Here, CR(X ｉ , Y ｉ ) is the contribution rate of factor X of the i-th watershed unit to disaster data Y, where 1 ≤ i ≤ bsn, and bsn is the number of watershed units in the target area, and X ｉ is the factor matrix of the i-th watershed unit, and X ｉ ∈X ｂｓｎ×ｃ×ｄｘ Here, dx is the number of factors, dx = 1 if there is a single factor, and if there are combination factors, dx = the number of combination factors, c is the number of events, Y ｉ ∈Y ｂｓｎ×ｃ×ｄｙ Here, dy is the number of indicators for disaster data, dy = 5, and σ(・) ２ The prediction and forecasting method for mountain tsunamis with various generation mechanisms as described in 1, characterized in that is the overall dispersion.

5. The specific process for establishing the water collection topology for each type of landslide described in Step 2 is characterized in that it starts from the source area, analyzes in the order of water collection from the upstream to the downstream basin, and if the type of landslide described as Category A of one basin unit does not match the type of landslide described as Category B of a downstream basin unit, the downstream type of Category A is determined to be Category B, and the upstream type of Category B is determined to be Category A, and if the two match, Category A and Category B are merged into one virtual category, and the search continues in the downstream basin units until a downstream basin unit that does not match the landslide type category is found, all basin units are visited until the basin unit at the outlet of the target area is reached, and finally the upstream and downstream type of each type of landslide is determined, thus providing a prediction and forecasting method suitable for landslides with various generation mechanisms as described in 1.

6. The specific process for normalizing the input / output interface and model structure for each established model described in Step 3 is as follows: For station-scale input data, normalization is performed in the data format of station code, station longitude, station latitude, type step, precipitation observed at the station, weather, snowmelt, and sediment load; for precipitation quantitatively estimated by radar and precipitation data retrieved from satellite data, normalization is performed according to the fourth type of grid data format of MICAPS; for gully measurement data, normalization is performed in the data format of cross section code, longitude, latitude, and elevation; and for the output short-term forecast precipitation data... The prediction and forecasting method for mountain tsunamis of various generation mechanisms according to claim 1, characterized in that: normalization processing is performed according to the fourth type of grid data format of MICAPS; the output gully flow process data is normalized in the form of model time step, current time step, model code, gully code, flow rate, water level, snowmelt amount, sediment load, and inundation depth data; the model file is normalized with basic information, parameter information, and model information; a unified input / output interface is adopted to generate a dynamic integrated framework for various types of models; and the model is registered / deployed, encapsulated / released, and called.

7. The specific process for constructing an integrated prediction and forecasting model for multi-type landslides in the target area described in Step 3 is as follows: For landslides with multiple upstream type classifications, the inundation depth process of all upstream watershed classifications is used as the upper boundary of the landslides with multiple upstream type classifications, and the sum of the calculation results of the remaining models excluding the inundation depth of all upstream type classifications is used as the input value for the gully of the landslides with multiple downstream type classifications; for landslides with multiple downstream type classifications, the inundation depth process of the landslides with multiple downstream type classifications is used as the lower boundary of all downstream type classifications, and the output value of the remaining gully excluding the inundation depth of the landslides with multiple downstream type classifications is determined according to the weight of the water-carrying cross-sectional area of ​​the gully of the downstream type classification; this is the prediction and forecasting method for landslides with various generation mechanisms described in 1.

8. The specific process for determining the optimal parameter set for the multi-type landslide prediction and forecasting integrated model described in Step 4 is as follows: Optimize the parameters of the landslide prediction and forecasting model for the heavy rain type landslide category by adopting the flow rate process data collected in Step 1; optimize the parameters of the landslide prediction and forecasting model for the snowmelt type and rain-snow combined type landslide category by adopting the flow rate and snowmelt process data; optimize the parameters of the landslide prediction and forecasting model for the water flow / sediment type landslide category by adopting the flow rate, water level, and sediment load process data and flood trace data; The prediction and forecasting method for mountain tsunamis with various generation mechanisms, as described in 1, is characterized by the following steps: optimizing the parameters of a mountain tsunami prediction and forecasting model for snow-sedipal composite type mountain tsunami classifications by employing distance data and flood trace data; determining the optimal parameters for the model corresponding to each classification; determining the parameters of mountain tsunami classification models where the data observed at observation stations and flood trace data are insufficient, based on the optimal parameters of each classification model determined by the following steps: determining the optimal parameter set for the integrated model using the mountain tsunami type classification as a unit.

9. The prediction and forecasting method for mountain tsunamis with various generation mechanisms, as described in 8, characterized in that the method for localization analysis of the parameters is a spatial similarity method, an attribute similarity method, a parameter regression method, or a classification tree / regression tree method.

Images