A flash flood prediction and forecasting method suitable for different genesis mechanisms

Abstract

This invention discloses a method for predicting and forecasting flash floods suitable for different formation mechanisms, comprising the following steps: Step 1, collecting and organizing basic data of the study area; Step 2, determining flash flood type zones and their catchment topology; Step 3, constructing an integrated model for predicting and forecasting multiple types of flash floods; Step 4, determining the optimal parameter set and evaluation index of the integrated model; Step 5, forecasting the flash flood process of each flash flood type zone and its outlet. The method described in this invention considers the differences in the formation mechanisms of flash floods, proposes an integrated model for predicting and forecasting multiple types of flash floods with physical mechanisms, and its parameter optimization strategy and evaluation index. It extends traditional single-type flash flood prediction and forecasting to multi-type flash flood prediction and forecasting, and can provide the main controlling factors and prediction and forecasting processes of flash flood formation, development, and disaster-causing processes in any region. It accurately predicts the physical processes of the formation and development of different types of flash floods, has strong applicability, and can better support the formulation of flash flood disaster prevention decisions.

Description

Technical Field

[0001] This invention belongs to the field of hydrological and meteorological forecasting technology, and in particular relates to a method for predicting and forecasting flash floods suitable for different formation mechanisms. Background Technology

[0002] Flash floods are the most severe type of flood disaster in my country, causing the most casualties and economic losses. They are characterized by their numerous locations, wide distribution, sudden onset, high mortality rate, and significant regional variations. With the increasing frequency of extreme precipitation events, the risk of flash floods continues to rise. Therefore, accurate forecasting and prediction of flash floods is a crucial technical means to improve flash flood prevention capabilities. However, due to the dual nature of flash floods—both natural and social—their formation, development, and damage are the result of the combined effects of multiple factors, including rainfall, underlying surface conditions, and human activities. These factors exhibit high spatiotemporal heterogeneity, leading to complex mechanisms of flash flood formation and evolution in different regions and small watersheds. The causes, destructiveness, and destructiveness of different types of flash floods also vary. Ignoring these differences in flash flood formation mechanisms inevitably leads to uncertainty in flash flood prediction.

[0003] Currently, single hydrological or hydrodynamic models are mainly used for simulating and forecasting flash floods in small watersheds, focusing primarily on single types of flash flood events, such as those caused by short-duration heavy rainfall. However, recent flash flood disasters have demonstrated the significant chain reactions and complex effects between processes such as rapid flood rise, sediment erosion, and glacial meltwater in small watersheds, leading to frequent major disasters caused by minor rainfall. Therefore, forecasting methods targeting single types of flash floods fail to consider the differences in their formation mechanisms, making it difficult to accurately reproduce the occurrence and development of other types of flash floods. This results in low accuracy in flash flood forecasting, severely impacting decision-making for flash flood disaster prevention. Therefore, there is an urgent need to develop flash flood forecasting methods suitable for different formation mechanisms to improve forecast accuracy and reduce flash flood risks and losses. Summary of the Invention

[0004] The purpose of this invention is to provide a method for predicting and forecasting flash floods suitable for different formation mechanisms, so as to solve the above-mentioned technical problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] This invention discloses a method for predicting and forecasting flash floods suitable for different formation mechanisms, the method comprising the following steps:

[0007] Step 1: Collect and organize basic data for the study area: Collect and organize various basic data for the study area, including multi-source precipitation, water level processes, flow processes, sediment load processes, snowmelt processes, meteorological data, distribution of villages along the river, gully measurements, basic geographic information, and historical flash flood disaster survey data; the multi-source precipitation data includes station-observed precipitation, radar quantitatively estimated precipitation, and satellite-retrieved precipitation data; the historical flash flood disaster survey data includes the number of deaths / missing persons caused by the disaster, the number of collapsed houses, direct economic losses, floodmark data, and inundated area; the floodmark data includes the location of floodmarks, the occurrence of floodmarks... The study determined the peak flood level, peak flood flow, and peak sediment load for each flood event by extracting the maximum values ​​from the water level, flow, and sediment load processes. The total snowmelt amount for each event was then determined by summing the snowmelt data. Based on the basic geographic information data of the study area and the distribution data of villages along the river, small watershed basic calculation units were divided according to a set area threshold, and the water system distribution and its catchment topology were extracted. Based on the divided small watershed basic calculation units, watershed units were further divided according to the set catchment area threshold and the distribution data of villages along the river, and the catchment topology of each watershed unit was established.

[0008] Step 2: Determine the flash flood type zones and their catchment topology: First, based on the peak flood level, peak flood flow, peak sediment load, total snowmelt, and historical flash flood disaster data for each event in the study area, analyze the main controlling factors of flash flood formation, development, and disaster-causing processes in each watershed; then, based on the main controlling factors and their values ​​for each watershed unit, use a clustering algorithm to determine the flash flood type zones in the study area; the flash flood type zones include rainstorm-type flash flood zones, snowmelt-type flash flood zones, water-sediment-type flash flood zones, rain-snow composite flash flood zones, and snow-sand composite flash flood zones; finally, based on the catchment topology of each watershed unit and the determined flash flood type zones, establish the catchment topology of each flash flood type zone;

[0009] Step 3: Construct an integrated model for multi-type flash flood prediction and forecasting: Set the model time step to hours or minutes, and the model spatial scale to small watershed basic computational units. Establish a near-term numerical precipitation forecasting model, a nonlinear runoff generation and confluence model, a snowmelt runoff generation and confluence model, and a non-uniform flow flash flood and sediment one-dimensional hydrodynamic model, and determine the input and output data for each model. The input data for the near-term numerical precipitation forecasting model are radar quantitatively estimated precipitation and satellite-retrieved precipitation data, and the output data is short-term forecast precipitation data. The input data for the nonlinear runoff generation and confluence model are station-observed precipitation and meteorological data, and the output data is channel flow process data. The input data for the snowmelt runoff generation and confluence model are station-observed precipitation and meteorological data, and the output data is channel flow and snowmelt process data. The input data for the non-uniform flow flash flood and sediment one-dimensional hydrodynamic model are station-observed precipitation and channel measurement data, and the output data are channel flow, water level, sediment load, and inundation depth process data.

[0010] The input / output interfaces and model structures of the established models were standardized to construct a comprehensive flash flood model library. Based on the determined flash flood type zones, corresponding computational models were selected from the comprehensive flash flood model library. These models were combined according to the sequence of precipitation-runoff-flood evolution in small watersheds to establish flash flood prediction and forecasting models for each flash flood type zone. Then, based on the output data of the flash flood prediction and forecasting models for each zone, the maximum values ​​were extracted from the gully water level process, flow process, sediment load process, and inundation depth process to determine the predicted peak flood level, peak flow, peak sediment load, and maximum inundation depth for each zone. The snowmelt amount in the gully was summed to determine the total predicted snowmelt amount for each zone. Based on the established catchment topology of each flash flood type zone, the upstream and downstream type zone models of the flash flood prediction and forecasting models for each zone were determined, constructing an integrated multi-type flash flood prediction and forecasting model for the study area.

[0011] Step 4: Determine the optimal parameter set and evaluation index of the integrated model: Based on the established catchment topology of each flash flood type zone, and in the order of upstream to downstream, calibrate the flash flood prediction and forecasting model of each flash flood type zone one by one, and determine the optimal parameter set of the integrated model for multi-type flash flood prediction and forecasting.

[0012] The accuracy of the multi-type flash flood prediction and forecasting integrated model and the flash flood prediction and forecasting models of each region were evaluated using a weighted comprehensive index. The optimal value was 1, and the calculation formula was as follows:

[0013] (1)

[0014] In the formula, g is the weighted comprehensive index of the multi-type flash flood forecasting and prediction integrated model; m is the number of flash flood type zones in the study area; i is the i-th flash flood type zone in the study area with station observation data and flood trace point data, 1≤i≤m; gi and α i These are the weighted composite indicators and their weights for the i-th partition. RMSE i Let be the root mean square error of precipitation in the i-th region. ;P o,j,i and P s,j,i M represents the observed precipitation at the station and the short-term forecast precipitation output by the model at time j in the i-th partition, respectively; i The sequence length of the precipitation data observed at the stations in the i-th partition; |Re i | represents the absolute value of the relative error of peak flow, peak water level, peak sediment load, snowmelt amount, or inundation depth for the i-th partition. Q o,p,i For the i-th partition, the peak flow rate, peak water level, peak sediment load, total snowmelt, or elevation of the flood mark point is determined by station observation data; Q s,p,i For the model predicting the peak flow, peak water level, peak sediment load, total snowmelt, or maximum inundation depth for the i-th partition; NSE i Let be the Nash efficiency coefficient of the flow process line, water level process line, sediment load process line, or snowmelt process line for the i-th partition. Q o,j,i For the station observation flow rate, water level, sediment load, or snowmelt at time j in the i-th partition; Q s,j,i For the model prediction of flow rate, water level, sediment load, or snowmelt at time j in the i-th partition; Let N be the average observed flow rate, average water level, average sediment load, or average snowmelt amount for the i-th partition; i The sequence length of the observed flow rate, water level, sediment load, or snowmelt amount for the i-th partition; β represents the weight of the precipitation assessment index for the i-th partition; i The weights of the evaluation indicators for the i-th partition are: flow rate, water level, sediment load, snowmelt, or inundation depth.

[0015] Step 5: Forecast the flash flood process for each flash flood type zone and its outlet: Based on the multi-source precipitation data, meteorological data, and gully measurement data collected in Step 1, drive the multi-type flash flood prediction and forecasting integrated model to forecast the short-term forecast precipitation, gully flow, water level, sediment load, snowmelt amount, and inundation depth for each flash flood type zone and its outlet in the study area.

[0016] Furthermore, the water level, flow rate, sediment load, and snowmelt data mentioned in step 1 are all station observation data; the meteorological data includes station observed air temperature, soil moisture, water surface evaporation, sunshine duration, and wind speed data; the gully measurement data includes cross-section, longitudinal section, and bridge / culvert measurement data; the basic geographic information data includes a DEM at a scale of no less than 1:50,000, land use type vector data, and soil texture type vector data; and the set area threshold is 50 km². 2 The set catchment area threshold is 200 km². 2 .

[0017] Furthermore, the specific process of establishing the water catchment topology of each watershed unit in step 1 is as follows: taking the watershed unit where the river source is located as the source watershed, and determining the river that the river flows into at the outlet of the source watershed according to the water catchment topology, the watershed unit where the river is located is the downstream watershed of the source watershed, and repeating the above process until there is no river flowing into the outlet of the watershed unit.

[0018] Furthermore, the specific process of analyzing the main controlling factors of the formation, development, and disaster-causing processes of flash floods in each watershed, as described in step 2, is as follows: Using multivariate data mining techniques, the contribution rates of single factors such as peak flood level, peak flood flow, peak sediment load, and total snowmelt amount for each flash flood event in each watershed unit to the corresponding number of deaths / missing persons, number of collapsed houses, direct economic losses, maximum flood mark elevation, and maximum inundated area disaster data are analyzed. Additionally, the contribution rates of multiple watershed combination factors for each flash flood event in each watershed unit to the corresponding disaster data are analyzed, including combinations of peak flood flow and total snowmelt amount, combinations of peak flood flow, peak flood level, and peak sediment load, and combinations of total snowmelt amount, peak flood level, and peak sediment load. If the contribution rate of a single watershed factor exceeds 0.5, then the single watershed factor is the main controlling factor of the formation, development, and disaster-causing processes of flash floods in each watershed. If the contribution rate of watershed combination factors is greater than the sum of the contribution rates of single factors, then the watershed combination factor is the main controlling factor of flash floods in the watershed.

[0019] The formula for calculating the contribution rate is:

[0020] (2)

[0021] In the formula, CR(X) i ,Y i Let X be the contribution rate of factor X of the i-th watershed unit to the disaster data Y, 1≤i≤bsn, where bsn is the number of watershed units in the study area; i Let X be the factor matrix of the i-th watershed unit. i ∈X bsn×c×dx dx represents the number of factors; if it is a single factor, dx = 1; if it is a combination factor, dx represents the number of combination factors; c represents the number of fields. i∈Y bsn×c×dy dy represents the number of disaster data indicators, dy=5; This represents the overall variance.

[0022] Furthermore, the specific process for establishing the catchment topology of each flash flood type zone in step 2 is as follows: Starting from the source basin, the analysis is performed according to the catchment order from the upstream basin to the downstream basin. If the flash flood type zone of a certain basin unit is denoted as zone A and the flash flood type zone of its downstream basin unit is denoted as zone B, then the downstream type zone of zone A is determined to be zone B, and the upstream type zone of zone B is determined to be zone A. If the two are consistent, then zone A and zone B are merged into a virtual zone, and the search continues to downstream basin units until a downstream basin unit that is inconsistent with its flash flood type zone is found. All basin units are traversed until the basin unit at the outlet of the study area, and finally the upstream and downstream type zones of each flash flood type zone are determined.

[0023] Furthermore, the specific process of standardizing the input / output interfaces and model structure of the established models described in step 3 is as follows: Station-scale input data is standardized according to the data formats of station code, station longitude, station latitude, time step, station observed precipitation, meteorological data, snowmelt data, and sediment load; radar quantitatively estimated precipitation and satellite-retrieved precipitation data are standardized according to the MICAPS Type 4 grid data format; channel measurement data are standardized according to the data formats of cross-section number, longitude, latitude, and elevation; output short-term forecast precipitation data are standardized according to the MICAPS Type 4 grid data format; output channel flow process data are standardized according to the data formats of model time step, current time step, model code, channel code, flow rate, water level, snowmelt data, sediment load, and inundation depth; model files are standardized according to basic information, parameter information, and model information; a unified input / output interface is used to form a dynamic integration framework for various models, enabling model registration, deployment, encapsulation, release, and invocation.

[0024] Furthermore, the specific process of constructing the integrated model for predicting and forecasting multiple types of flash floods in the study area as described in step 3 is as follows: For a flash flood type partition with multiple upstream type partitions, the inundation depth process of all upstream watershed partitions is taken as the upper boundary of the flash flood type partition, and the sum of the calculation results of all upstream type partitions except for the inundation depth is taken as the channel input value of the flash flood type partition. For a flash flood type partition with multiple downstream type partitions, the inundation depth process of the flash flood type partition is taken as the lower boundary of all downstream type partitions, and the channel output values ​​of the flash flood type partition except for the inundation depth are determined by the channel cross-sectional area weight of the downstream type partition.

[0025] Furthermore, the specific process for determining the optimal parameter set of the integrated model for multi-type flash flood prediction and forecasting in step 4 is as follows: The parameters for the flash flood prediction and forecasting model of the rainstorm-type flash flood zone are optimized using the flow process data collected in step 1; the parameters for the flash flood prediction and forecasting model of the snowmelt-type and rain-snow composite-type flash flood zone are optimized using flow and snowmelt process data; the parameters for the flash flood prediction and forecasting model of the water-sediment-type flash flood zone are optimized using flow, water level, and sediment load process data, as well as flood mark point data; and the parameters for the flash flood prediction and forecasting model of the snow-sand composite-type flash flood zone are optimized using flow, water level, snowmelt, and sediment load process data, as well as flood mark point data. The optimal parameters for the corresponding zones are determined respectively. Based on the determined optimal parameters for each zone, the parameter regionalization analysis method is used to determine the model parameters for flash flood zones lacking station observation data and flood mark point data. The optimal parameter set for the integrated model is determined using flash flood type zones as units.

[0026] Furthermore, the parameter regionalization analysis method is spatial proximity method, attribute similarity method, parameter regression method, or classification and regression number method.

[0027] The beneficial effects of this invention are as follows: The method described in this invention takes into account the differences in the formation mechanism of flash floods, and adopts a combination of multivariate data mining technology and multiple numerical simulation technology to propose an integrated model for prediction and forecasting of multiple types of flash floods with physical mechanisms, as well as its parameter optimization strategy and evaluation index. This extends the traditional single-type flash flood prediction and forecasting to multi-type flash flood prediction and forecasting, and can provide the main controlling factors and prediction and forecasting process of flash flood formation, development and disaster-causing processes in any region. It can accurately predict the physical processes of the formation and development of different types of flash floods, and the analysis is more comprehensive and applicable, which can better support the decision-making of flash flood disaster prevention.

[0028] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the method flow described in this invention;

[0030] Figure 2 A schematic diagram illustrating the process of determining flash flood type zones and their catchment topology;

[0031] Figure 3 A schematic diagram illustrating the construction process of an integrated model for predicting and forecasting multiple types of flash floods. Detailed Implementation

[0032] This invention discloses a method for predicting and forecasting flash floods suitable for different formation mechanisms, such as... Figures 1-3 As shown, the method includes the following steps:

[0033] Step 1: Collect and organize basic data for the study area.

[0034] Various basic data were collected and organized for the study area, mainly including multi-source precipitation, water level processes, flow processes, sediment load processes, snowmelt processes, meteorological data, distribution of riverside villages, gully measurements, basic geographic information, and historical flash flood disaster surveys. Maximum values ​​were extracted from the water level, flow, and sediment load processes for each event to determine the peak flood level, peak flow, and peak sediment load for each event. The total snowmelt amount for each event was determined by summing the snowmelt data. Multi-source precipitation data included station-observed precipitation, radar quantitatively estimated precipitation, and satellite-retrieved precipitation data. Water level, flow, and... The data on sediment load and snowmelt are all from station observations; meteorological data include station-observed air temperature, soil moisture, water surface evaporation, sunshine hours, and wind speed; gully measurement data includes cross-section, longitudinal section, and bridge and culvert measurement data; basic geographic information data includes DEMs at a scale of no less than 1:50,000, land use type vector data, and soil texture type vector data; historical flash flood disaster survey data includes the number of deaths / missing persons, number of collapsed houses, direct economic losses, floodmark data, and inundated area caused by the disaster; floodmark data includes the location of floodmarks, the time of occurrence of floodmarks, and the elevation of floodmarks.

[0035] Then, based on the basic geographic information data of the study area and the distribution data of villages along the river, a 50 km... 2 Using area thresholds, small watersheds are divided into basic computational units, and the distribution of water systems and their catchment topology are extracted; based on the divided small watershed basic computational units, according to 200 km... 2 Based on the water catchment area threshold and the distribution data of villages along the river, watershed units are divided, and the water catchment topology of each watershed unit is established. Specifically, the watershed unit where the river source is located is taken as the source watershed. According to the water catchment topology, the river that flows into the river at the outlet of the source watershed is determined. The watershed unit where this river is located is the downstream watershed of the source watershed. The above process is repeated until there is no river flowing into the river at the outlet of the watershed unit.

[0036] Step 2: Determine the flash flood type zones and their catchment topology.

[0037] First, based on the peak flood level, peak flood flow, peak sediment load, total snowmelt, and historical flash flood disaster data for each event in the study area, the main controlling factors of flash flood formation, development, and disaster-causing processes in each watershed are analyzed. The specific process is as follows: Using multivariate data mining techniques such as geographic detectors and sorting analysis, the contribution rates of single factors (peak water level, peak flow, peak sediment load, and total snowmelt) to disaster data such as the number of deaths / missing persons, number of collapsed houses, direct economic losses, maximum flood mark elevation, and maximum inundated area for each flash flood event in each watershed unit are analyzed. The contribution rates of multiple watershed combination factors to the corresponding disaster data for each flash flood event in each watershed unit are also analyzed, including combinations of peak flow and total snowmelt, combinations of peak flow, peak water level, and peak sediment load, and combinations of total snowmelt, peak water level, and peak sediment load. If the contribution rate of a single watershed factor exceeds 0.5, then the single watershed factor is the dominant factor in the formation, development, and disaster-causing process of flash floods in each watershed. If the contribution rate of watershed combination factors is greater than the sum of the contribution rates of single factors, then the watershed combination factors are the dominant factors in the flash floods of the watershed. The contribution rate calculation formula is shown below:

[0038] (2)

[0039] In the formula, CR(X) i ,Y i Let X be the contribution rate of factor X of the i-th watershed unit to the disaster data Y, 1≤i≤bsn, where bsn is the number of watershed units in the study area; i Let X be the factor matrix of the i-th watershed unit. i ∈X bsn×c×dx dx represents the number of factors; if it is a single factor, dx = 1; if it is a combination factor, dx represents the number of combination factors; c represents the number of fields. i ∈Y bsn×c×dy dy represents the number of disaster data indicators, dy=5; This represents the overall variance.

[0040] Then, based on the main controlling factors and their values ​​for each watershed unit, a clustering algorithm was used to determine the main flash flood types and their zones in the study area. The main flash flood types include: rainstorm type, snowmelt type, water-sediment type, rain-snow composite type, and snow-sand composite type. The corresponding flash flood type zones are: rainstorm type flash flood zone, snowmelt type flash flood zone, water-sediment type flash flood zone, rain-snow composite type flash flood zone, and snow-sand composite type flash flood zone. Based on the catchment topology of each watershed unit and the determined flash flood type zones, the catchment topology of each flash flood type zone was established, specifically: starting from the source watershed, following the pattern from upstream to downstream... The water catchment sequence of the upstream watershed is analyzed. If the flash flood type partition (denoted as partition A) of a certain watershed unit is inconsistent with the flash flood type partition (denoted as partition B) of its downstream watershed unit, then the downstream type partition of partition A is determined to be partition B, and the upstream type partition of partition B is determined to be partition A. If the two are consistent, then partition A and partition B are merged into a virtual partition, and the search continues to downstream watershed units until a downstream watershed unit inconsistent with its flash flood type partition is found. All watershed units are traversed until the watershed unit at the outlet of the study area, and finally the upstream and downstream type partitions of each flash flood type partition are determined.

[0041] Step 3: Construct an integrated model for predicting and forecasting multiple types of flash floods.

[0042] The model time step is set to hours or minutes, and the model spatial scale is the basic computational unit of a small watershed. A near-term numerical precipitation forecast model, a nonlinear runoff generation model, a snowmelt runoff generation model, and a non-uniform mountain flood-sediment 1 / 2 hydrodynamic model are established, and the input and output data for each model are determined. Specifically, the input data for the near-term numerical precipitation forecast model consists of radar quantitatively estimated precipitation and satellite-retrieved precipitation data, and the output data is the near-term forecast precipitation data. The input data for the nonlinear runoff generation model consists of station-observed precipitation and meteorological data, and the output data is channel flow process data. The input data for the snowmelt runoff generation model consists of station-observed precipitation and meteorological data, and the output data is channel flow and snowmelt process data. The input data for the non-uniform mountain flood-sediment 1 / 2 hydrodynamic model consists of station-observed precipitation and channel measurement data, and the output data is channel flow, water level, sediment load, and inundation depth process data.

[0043] The input / output interfaces and model structures of the established models are standardized. Specifically, station-scale input data are standardized according to the data formats of station code, station longitude, station latitude, time step, station observed precipitation, meteorological data, snowmelt data, and sediment load. Radar quantitatively estimated precipitation and satellite-retrieved precipitation data are standardized according to the MICAPS Type 4 grid data format. Gully measurement data are standardized according to the data formats of cross-section number, longitude, latitude, and elevation. Output short-term forecast precipitation data are standardized according to the MICAPS Type 4 grid data format. Output gully flow process data are standardized according to the data formats of model time step, current time step, model code, gully code, flow rate, water level, snowmelt data, sediment load, and inundation depth. Model files are standardized according to basic information, parameter information, and model information. A unified input / output interface is used to form a dynamic integration framework for various models, enabling model registration, deployment, encapsulation, release, and invocation. This leads to the construction of a comprehensive flash flood model library.

[0044] Based on the identified flash flood type zones, appropriate calculation models were selected from the full-type flash flood model library. These models were combined according to the sequence of precipitation-runoff-flood evolution in small watersheds to establish flash flood prediction and forecasting models for each flash flood type zone, as shown in Table 1. Then, based on the output data of each zone's flash flood prediction and forecasting models, the maximum values ​​were extracted from the gully water level process, flow process, sediment load process, and inundation depth process to determine the predicted peak flood level, peak flow, peak sediment load, and maximum inundation depth for each zone. The predicted total snowmelt amount for each zone was then summed over the gully snowmelt process.

[0045] Table 1. List of calculation model combinations involved in different flash flood type zones

[0046]

[0047] Based on the established catchment topology of each flash flood type zone, the upstream and downstream type zone models of the flash flood prediction and forecasting model for each zone are determined. An integrated model for multi-type flash flood prediction and forecasting in the study area is constructed. The integration strategy is as follows: For a flash flood type zone with multiple upstream type zones, the inundation depth process of all upstream watershed zones is taken as the upper boundary of the flash flood type zone, and the sum of the calculation results of all upstream type zones except for the inundation depth is taken as the channel input value of the flash flood type zone. For a flash flood type zone with multiple downstream type zones, the inundation depth process of the flash flood type zone is taken as the lower boundary of all downstream type zones, and the channel output values ​​of the flash flood type zone except for the inundation depth are determined by the channel cross-sectional area weight of the downstream type zones.

[0048] Step 4: Determine the optimal parameter set and evaluation index for the integrated model.

[0049] Based on the established catchment topology of each flash flood type zone, and following the order of upstream to downstream, the flash flood prediction and forecasting models for each flash flood type zone were calibrated one by one, forming a parameter optimization strategy for the integrated model of multi-type flash flood prediction and forecasting in the study area. Specifically, the parameters of the flash flood prediction and forecasting model for the rainstorm-type flash flood zone were optimized using the station observation flow data collected in step 1; the parameters of the flash flood prediction and forecasting model for the snowmelt-type and rain-snow composite-type flash flood zones were optimized using the station observation flow, water level, and sediment load data, as well as flood mark data, were optimized. The parameters of the flash flood prediction and forecasting model for water-sediment type flash flood zones were optimized. Station observation data on flow, water level, snowmelt amount, and sediment load, as well as flood trace data, were used to optimize the parameters of the flash flood prediction and forecasting model for snow-sediment composite flash flood zones, determining the optimal model parameters for each zone. Based on the determined optimal model parameters for each zone, regionalization analysis methods such as spatial proximity, attribute similarity, parameter regression, or classification and regression number methods were used to determine the model parameters for flash flood zones lacking station observation data and flood trace data. An integrated model optimal parameter set was determined using flash flood type zones as units.

[0050] The accuracy of the ensemble model and the partitioned model is evaluated using a weighted comprehensive index, with the optimal value being 1. The calculation formula is as follows:

[0051] (1)

[0052] In the formula, g is the weighted comprehensive index of the multi-type flash flood forecasting and prediction integrated model; m is the number of flash flood type zones in the study area; i is the i-th flash flood type zone in the study area with station observation data and flood trace point data, 1≤i≤m; g i and α i These are the weighted composite indicators and their weights for the i-th partition. RMSE i Let be the root mean square error of precipitation in the i-th region. ;P o,j,i and P s,j,i M represents the observed precipitation at the station and the short-term forecast precipitation output by the model at time j in the i-th partition, respectively; i The sequence length of the precipitation data observed at the stations in the i-th partition; |Re i | represents the absolute value of the relative error of peak flow, peak water level, peak sediment load, snowmelt amount, or inundation depth for the i-th partition. Q o,p,iFor the i-th partition, the peak flow rate, peak water level, peak sediment load, total snowmelt, or elevation of the flood mark point is determined by station observation data; Q s,p,i For the model predicting the peak flow, peak water level, peak sediment load, total snowmelt, or maximum inundation depth for the i-th partition; NSE i Let be the Nash efficiency coefficient of the flow process line, water level process line, sediment load process line, or snowmelt process line for the i-th partition. Q o,j,i For the station observation flow rate, water level, sediment load, or snowmelt at time j in the i-th partition; Q s,j,i For the model prediction of flow rate, water level, sediment load, or snowmelt at time j in the i-th partition; Let N be the average observed flow rate, average water level, average sediment load, or average snowmelt amount for the i-th partition; i The sequence length of the observed flow rate, water level, sediment load, or snowmelt amount for the i-th partition; β represents the weight of the precipitation assessment index for the i-th partition; i The weights of the evaluation indicators for the i-th partition are: flow rate, water level, sediment load, snowmelt, or inundation depth.

[0053] Step 5: Forecast the flash flood process for each flash flood type zone and its outlet.

[0054] Based on the multi-source precipitation data, meteorological data, and gully measurement data collected in step 1, a multi-type flash flood prediction and forecasting integrated model is driven to predict the short-term forecast precipitation, gully flow, water level, sediment load, snowmelt amount, and inundation depth of each flash flood type zone and the outlet of the study area.

[0055] Finally, it should be noted that the above description is only used to illustrate the technical solution of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred arrangement, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for predicting and forecasting flash floods suitable for different formation mechanisms, characterized in that, The method includes the following steps: Step 1: Collect and organize basic data for the study area: Collect and organize various basic data for the study area, including multi-source precipitation, water level processes, flow processes, sediment load processes, snowmelt processes, meteorological data, distribution of villages along the river, gully measurements, basic geographic information, and historical flash flood disaster survey data; the multi-source precipitation data includes station-observed precipitation, radar quantitatively estimated precipitation, and satellite-retrieved precipitation data; the historical flash flood disaster survey data includes the number of deaths / missing persons caused by the disaster, the number of collapsed houses, direct economic losses, floodmark data, and inundated area; the floodmark data includes the location of floodmarks, the occurrence of floodmarks... The study determined the peak flood level, peak flood flow, and peak sediment load for each flood event by extracting the maximum values ​​from the water level, flow, and sediment load processes. The total snowmelt amount for each event was then determined by summing the snowmelt data. Based on the basic geographic information data of the study area and the distribution data of villages along the river, small watershed basic calculation units were divided according to a set area threshold, and the water system distribution and its catchment topology were extracted. Based on the divided small watershed basic calculation units, watershed units were further divided according to the set catchment area threshold and the distribution data of villages along the river, and the catchment topology of each watershed unit was established. Step 2: Determine the flash flood type zones and their catchment topology: First, based on the peak flood level, peak flood flow, peak sediment load, total snowmelt, and historical flash flood disaster data for each event in the study area, analyze the main controlling factors of flash flood formation, development, and disaster-causing processes in each watershed; then, based on the main controlling factors and their values ​​for each watershed unit, use a clustering algorithm to determine the flash flood type zones in the study area; the flash flood type zones include rainstorm-type flash flood zones, snowmelt-type flash flood zones, water-sediment-type flash flood zones, rain-snow composite flash flood zones, and snow-sand composite flash flood zones; finally, based on the catchment topology of each watershed unit and the determined flash flood type zones, establish the catchment topology of each flash flood type zone; Step 3: Construct an integrated model for multi-type flash flood prediction and forecasting: Set the model time step to hours or minutes, and the model spatial scale to small watershed basic computational units. Establish a near-term numerical precipitation forecasting model, a nonlinear runoff generation and confluence model, a snowmelt runoff generation and confluence model, and a non-uniform flow flash flood and sediment one-dimensional hydrodynamic model, and determine the input and output data for each model. The input data for the near-term numerical precipitation forecasting model are radar quantitatively estimated precipitation and satellite-retrieved precipitation data, and the output data is short-term forecast precipitation data. The input data for the nonlinear runoff generation and confluence model are station-observed precipitation and meteorological data, and the output data is channel flow process data. The input data for the snowmelt runoff generation and confluence model are station-observed precipitation and meteorological data, and the output data is channel flow and snowmelt process data. The input data for the non-uniform flow flash flood and sediment one-dimensional hydrodynamic model are station-observed precipitation and channel measurement data, and the output data are channel flow, water level, sediment load, and inundation depth process data. The input / output interfaces and model structures of the established models were standardized to construct a comprehensive flash flood model library. Based on the determined flash flood type zones, corresponding computational models were selected from the comprehensive flash flood model library. These models were combined according to the sequence of precipitation-runoff-flood evolution in small watersheds to establish flash flood prediction and forecasting models for each flash flood type zone. Then, based on the output data of the flash flood prediction and forecasting models for each zone, the maximum values ​​were extracted from the gully water level process, flow process, sediment load process, and inundation depth process to determine the predicted peak flood level, peak flow, peak sediment load, and maximum inundation depth for each zone. The snowmelt amount in the gully was summed to determine the total predicted snowmelt amount for each zone. Based on the established catchment topology of each flash flood type zone, the upstream and downstream type zone models of the flash flood prediction and forecasting models for each zone were determined, constructing an integrated multi-type flash flood prediction and forecasting model for the study area. Step 4: Determine the optimal parameter set and evaluation index of the integrated model: Based on the established catchment topology of each flash flood type zone, and in the order of upstream to downstream, calibrate the flash flood prediction and forecasting model of each flash flood type zone one by one, and determine the optimal parameter set of the integrated model for multi-type flash flood prediction and forecasting. The accuracy of the multi-type flash flood prediction and forecasting integrated model and the flash flood prediction and forecasting models of each region were evaluated using a weighted comprehensive index. The optimal value was 1, and the calculation formula was as follows: （1） In the formula, g is the weighted comprehensive index of the multi-type flash flood forecasting and prediction integrated model; m is the number of flash flood type zones in the study area; i is the i-th flash flood type zone in the study area with station observation data and flood trace point data, 1≤i≤m; g i and α i These are the weighted composite indicators and their weights for the i-th partition. RMSE i Let be the root mean square error of precipitation in the i-th region. ;P o,j,i and P s,j,i M represents the observed precipitation at the station and the short-term forecast precipitation output by the model at time j in the i-th partition, respectively; i The sequence length of the precipitation data observed at the stations in the i-th partition; |Re i | represents the absolute value of the relative error of peak flow, peak water level, peak sediment load, snowmelt amount, or inundation depth for the i-th partition. Q o,p,i For the i-th partition, the peak flow rate, peak water level, peak sediment load, total snowmelt, or elevation of the flood mark point is determined by station observation data; Q s,p,i For the model predicting the peak flow, peak water level, peak sediment load, total snowmelt, or maximum inundation depth for the i-th partition; NSE i Let be the Nash efficiency coefficient of the flow process line, water level process line, sediment load process line, or snowmelt process line for the i-th partition. Q o,j,i For the station observation flow rate, water level, sediment load, or snowmelt at time j in the i-th partition; Q s,j,i For the model prediction of flow rate, water level, sediment load, or snowmelt at time j in the i-th partition; Let N be the average observed flow rate, average water level, average sediment load, or average snowmelt amount for the i-th partition; i The sequence length of the observed flow rate, water level, sediment load, or snowmelt amount for the i-th partition; β represents the weight of the precipitation assessment index for the i-th partition; i The weights of the evaluation indicators for the i-th partition are: flow rate, water level, sediment load, snowmelt, or inundation depth. Step 5: Forecast the flash flood process for each flash flood type zone and its outlet: Based on the multi-source precipitation data, meteorological data, and gully measurement data collected in Step 1, drive the multi-type flash flood prediction and forecasting integrated model to forecast the short-term forecast precipitation, gully flow, water level, sediment load, snowmelt amount, and inundation depth for each flash flood type zone and its outlet in the study area.

2. The method for predicting and forecasting flash floods suitable for different formation mechanisms according to claim 1, characterized in that, The water level, flow rate, sediment load, and snowmelt data mentioned in Step 1 are all station observation data; the meteorological data includes station observed air temperature, soil moisture, water surface evaporation, sunshine duration, and wind speed data; the gully measurement data includes cross-section, longitudinal section, and bridge / culvert measurement data; the basic geographic information data includes a DEM at a scale of no less than 1:50,000, land use type vector data, and soil texture type vector data; the set area threshold is 50 km². 2 The set catchment area threshold is 200 km². 2 .

3. The method for predicting and forecasting flash floods suitable for different formation mechanisms according to claim 1, characterized in that, The specific process of establishing the catchment topology of each watershed unit in step 1 is as follows: taking the watershed unit where the river source is located as the source watershed, and determining the river that the river flows into at the outlet of the source watershed according to the catchment topology of the river system, the watershed unit where the river is located is the downstream watershed of the source watershed, and repeating the above process until there is no river flowing into the outlet of the watershed unit.

4. The method for predicting and forecasting flash floods suitable for different formation mechanisms according to claim 1, characterized in that, The specific process for analyzing the controlling factors of flash flood formation, development, and disaster-causing processes in each watershed, as described in step 2, is as follows: Using multivariate data mining techniques, the contribution rates of single factors such as peak flood level, peak flood flow, peak sediment load, and total snowmelt to the corresponding number of deaths / missing persons, number of collapsed houses, direct economic losses, maximum flood mark elevation, and maximum inundated area for each flash flood event in each watershed unit are analyzed. Additionally, the contribution rates of multiple watershed combination factors to the corresponding disaster data for each flash flood event in each watershed unit are analyzed, including combinations of peak flood flow and total snowmelt, combinations of peak flood flow, peak flood level, and peak sediment load, and combinations of total snowmelt, peak flood level, and peak sediment load. If the contribution rate of a single watershed factor exceeds 0.5, then the single watershed factor is the controlling factor of the flash flood formation, development, and disaster-causing processes in each watershed. If the contribution rate of the combined watershed factors is greater than the sum of the contribution rates of the single factors, then the combined watershed factors are the controlling factors of the flash floods in the watershed. The formula for calculating the contribution rate is: （2） In the formula, CR(X) i ,Y i Let X be the contribution rate of factor X of the i-th watershed unit to the disaster data Y, 1≤i≤bsn, where bsn is the number of watershed units in the study area; i Let X be the factor matrix of the i-th watershed unit. i ∈X bsn×c×dx dx represents the number of factors; if it is a single factor, dx = 1; if it is a combination factor, dx represents the number of combination factors; c represents the number of fields. i ∈Y bsn×c×dy dy represents the number of disaster data indicators, dy=5; This represents the overall variance.

5. The method for predicting and forecasting flash floods suitable for different formation mechanisms according to claim 1, characterized in that, The specific process for establishing the catchment topology of each flash flood type zone in step 2 is as follows: Starting from the source basin, the analysis is carried out according to the catchment order from the upstream basin to the downstream basin. If the flash flood type zone of a certain basin unit is denoted as zone A and the flash flood type zone of its downstream basin unit is denoted as zone B, then the downstream type zone of zone A is determined to be zone B, and the upstream type zone of zone B is determined to be zone A. If the two are consistent, then zone A and zone B are merged into a virtual zone, and the search continues to the downstream basin units until a downstream basin unit that is inconsistent with its flash flood type zone is found. By traversing all watershed units up to the watershed unit at the outlet of the study area, the upstream and downstream type subdivisions of each flash flood type subdivision were finally determined.

6. The method for predicting and forecasting flash floods suitable for different formation mechanisms according to claim 1, characterized in that, The specific process of standardizing the input / output interfaces and model structure of the established models in step 3 is as follows: Station-scale input data is standardized according to the data formats of station code, station longitude, station latitude, time step, station observed precipitation, meteorological data, snowmelt data, and sediment load. Radar quantitatively estimated precipitation and satellite-retrieved precipitation data are standardized according to the MICAPS Type 4 grid data format. Channel measurement data are standardized according to the data formats of cross-section number, longitude, latitude, and elevation. Output short-term forecast precipitation data is standardized according to the MICAPS Type 4 grid data format. Output channel flow process data is standardized according to the data formats of model time step, current time step, model code, channel code, flow rate, water level, snowmelt data, sediment load, and inundation depth. Model files are standardized according to basic information, parameter information, and model information. A unified input / output interface is used to form a dynamic integration framework for various models, enabling model registration, deployment, encapsulation, release, and invocation.

7. A method for predicting and forecasting flash floods suitable for different formation mechanisms, as described in claim 1, is characterized in that... The specific process of constructing the integrated model for predicting and forecasting multiple types of flash floods in the study area as described in step 3 is as follows: For a flash flood type partition with multiple upstream type partitions, the inundation depth process of all upstream watershed partitions is taken as the upper boundary of the flash flood type partition, and the sum of the calculation results of all upstream type partitions except for the inundation depth is taken as the channel input value of the flash flood type partition. For a flash flood type partition with multiple downstream type partitions, the inundation depth process of the flash flood type partition is taken as the lower boundary of all downstream type partitions, and the channel output values ​​of the flash flood type partition except for the inundation depth are determined by the channel cross-sectional area weight of the downstream type partition.

8. A method for predicting and forecasting flash floods suitable for different formation mechanisms, as described in claim 1, is characterized in that... The specific process for determining the optimal parameter set of the integrated model for multi-type flash flood prediction and forecasting in step 4 is as follows: The parameters for the flash flood prediction and forecasting model of the rainstorm-type flash flood zone are optimized using the flow process data collected in step 1; the parameters for the flash flood prediction and forecasting model of the snowmelt-type and rain-snow composite-type flash flood zone are optimized using flow and snowmelt process data; the parameters for the flash flood prediction and forecasting model of the water-sediment-type flash flood zone are optimized using flow, water level, and sediment load process data, as well as flood mark point data; and the parameters for the flash flood prediction and forecasting model of the snow-sand composite-type flash flood zone are optimized using flow, water level, snowmelt, and sediment load process data, as well as flood mark point data. The optimal parameters for the corresponding zones are determined respectively. Based on the determined optimal parameters for each zone, the parameter regionalization analysis method is used to determine the model parameters for flash flood zones lacking station observation data and flood mark point data. The optimal parameter set for the integrated model is determined using flash flood type zones as units.

9. A method for predicting and forecasting flash floods suitable for different formation mechanisms, as described in claim 8, is characterized in that... The parameter regionalization analysis method is spatial proximity method, attribute similarity method, parameter regression method, or classification and regression number method.

Images