Personnel, traffic and economy oriented multi-dimensional flood risk assessment method and system

Abstract

The application discloses a multi-dimensional flood risk assessment method and system for personnel, traffic and economy. The method comprises the following steps: acquiring multi-source data, generating flood space-time evolution data through hydrodynamic simulation; constructing a time-dependent traffic network driven by flood dynamics, wherein the road traffic capacity dynamically decays with the water dynamic parameters; based on the space-time racing mechanism, calculating the time difference between the dangerous arrival time of personnel and the shortest evacuation time, constructing the evacuation time margin and the evacuation success probability model, dynamically correcting the static personnel exposure risk, and obtaining the personnel safety risk index; combining the physical state decay and asset exposure characteristics, respectively calculating the traffic operation and economic loss risk indexes, and performing multi-dimensional fusion evaluation. The application solves the problem that the traditional static assessment ignores the dynamic interaction between flood evolution and personnel evacuation, realizes the accurate calculation of the personnel trapped probability and the marginal contribution of key road sections, and provides scientific decision support for flood control command and emergency rescue.

Description

Technical Field

[0001] This invention belongs to the field of disaster risk assessment technology, especially a multi-dimensional flood risk assessment method and system for personnel, transportation, and economy. Background Technology

[0002] Traditional flood disaster assessment methods mostly focus on single areas, such as human safety, traffic disruption, or economic losses, lacking means to comprehensively assess disaster risks from multiple dimensions. In practical applications, the impacts of flood disasters are usually multifaceted and intertwined, with human exposure, traffic disruption, and economic losses influencing each other. A single assessment method cannot fully and accurately reflect the entirety of the disaster, resulting in inaccurate risk assessment results that are difficult to provide effective support for relevant decision-making.

[0003] Currently, research on flood disaster risk assessment technology mainly focuses on hydrodynamic simulation and loss assessment models. Hydrodynamic simulation, by comprehensively considering factors such as rainfall, watershed characteristics, and topography, can accurately predict the inundation range and depth of floods, providing basic data for disaster emergency response. However, hydrodynamic models still have certain limitations in assessing personnel risk, traffic impact, and economic losses. Traditional risk assessment methods often neglect the vulnerability of different socioeconomic groups, the connectivity analysis of transportation networks, and complex economic impacts, making it difficult to comprehensively consider the multidimensional risks of flood disasters.

[0004] To address this issue, risk assessment methods based on complex network theory and intelligent algorithms have received widespread attention in recent years. Complex network theory can be used to analyze the connectivity of transportation networks and identify the impact of traffic disruptions on the overall transportation system. Human instability mechanisms and multi-criteria decision-making methods provide new technical pathways for calculating personnel exposure and vulnerability. Intelligent algorithms (such as machine learning and fuzzy comprehensive evaluation methods) can fuse and analyze multi-source data, effectively improving the accuracy and intelligence of risk assessment. However, in practical applications, existing technologies suffer from the following key problems: First, existing methods cannot comprehensively assess the multi-dimensional risks to personnel, transportation, and the economy, leading to inaccurate assessment results; second, traditional assessment methods fail to fully consider the dynamic evolution of floods and their spatiotemporal impacts on personnel, transportation, and the economy; third, the impact of floods is multi-faceted, and existing methods lack effective comprehensive analysis, failing to accurately reflect the interactions between various dimensions.

[0005] Therefore, existing methods cannot meet practical needs, and a multidimensional risk assessment system is needed that can comprehensively assess flood disaster risks, has spatiotemporal dynamic analysis capabilities, and can respond quickly. Summary of the Invention

[0006] The purpose of this invention is to provide a multi-dimensional flood risk assessment method and system that considers personnel, transportation, and the economy, in order to solve one of the problems existing in the prior art.

[0007] The technical solution, a multi-dimensional flood risk assessment method considering personnel, transportation, and the economy, includes:

[0008] Acquire multi-source basic geographic data for the target area. Multi-source basic geographic data should include at least hydrological and meteorological data, topographic and geomorphological data, population distribution data, transportation network data, and economic asset data.

[0009] Based on the hydrodynamic model, the evolution simulation of multi-source basic geographic data is carried out to generate flood spatiotemporal evolution data, which includes the time-varying inundation depth sequence and water flow velocity sequence.

[0010] Based on flood spatiotemporal evolution data and population distribution data, the interaction characteristics between flood intensity and vulnerability of disaster-bearing bodies are analyzed, and the personnel safety risk index is calculated.

[0011] Based on flood spatiotemporal evolution data and traffic network data, we analyze the attenuation characteristics of the physical state of traffic facilities and network capacity, and calculate the traffic operation risk index.

[0012] Based on flood spatiotemporal evolution data and economic asset data, we analyze the exposure degree and value loss rate of different asset types and calculate the economic loss risk index.

[0013] A multi-dimensional integrated evaluation was conducted based on the personnel safety risk index, traffic operation risk index, and economic loss risk index to obtain the multi-dimensional comprehensive flood risk assessment results for the target area.

[0014] Beneficial effects: This invention solves the problem that traditional static assessments ignore the dynamic interaction between flood evolution and personnel evacuation, and enables accurate calculation of the probability of people being trapped and the marginal contribution of clearing key road sections, providing scientific decision support for flood control command and emergency rescue. Attached Figure Description

[0015] Figure 1 This is a diagram illustrating the overall technical process and system architecture of a multi-dimensional flood risk assessment targeting personnel, transportation, and the economy, as described in this application.

[0016] Figure 2 This is a flowchart illustrating the steps for calculating the personnel safety risk index in this application embodiment.

[0017] Figure 3 This is a flowchart illustrating the steps for calculating the traffic operation risk index in this embodiment of the application.

[0018] Figure 4 This is a flowchart illustrating the steps for calculating the economic loss risk index in this application embodiment.

[0019] Figure 5 This is a flowchart illustrating the steps for calculating the traffic operation risk index in this embodiment of the application. Detailed Implementation

[0020] To solve these problems, combined with Figures 1 to 5 The present invention will be specifically described through the following embodiments.

[0021] Example 1 describes the overall technical process and system architecture for multi-dimensional flood risk assessment focusing on personnel, transportation, and the economy. This example constructs a general framework compatible with both static and dynamic assessments, capable of flexibly configuring assessment modes according to different data accuracies and emergency needs.

[0022] Step 101: Obtain multi-source basic geographic data for the target area. The multi-source basic geographic data shall include at least hydrological and meteorological data, topographic and geomorphological data, population distribution data, transportation network data, and economic asset data.

[0023] In this embodiment, acquiring multi-source basic geographic data is a prerequisite for constructing a high-precision risk assessment model. Specifically, hydrological and meteorological data can include time-series rainfall data, evaporation, wind speed and direction, and historical flood records. These data typically originate from meteorological bureau monitoring stations or inversion products from meteorological satellites. Topographic data is preferably a high-resolution Digital Elevation Model (DEM), such as raster data stored in GeoTIFF format, with a spatial resolution of 30 meters, 10 meters, or even higher, used to accurately depict surface undulations and river morphology. Population distribution data can include resident population data from administrative divisions in statistical yearbooks, or dynamic heat map data based on location-based services (LBS). Transportation network data is typically represented as a vector topology map, including road centerlines, node coordinates, road classifications (such as expressways, arterial roads, and secondary arterial roads), and attribute information for bridges and tunnels. Economic asset data includes land use type maps, estimated housing values ​​for different regions, and asset statistical reports of key industrial and mining enterprises. Before these heterogeneous data enter the system, they need to undergo unified coordinate system registration (such as conversion to the WGS84 coordinate system) and data cleaning to ensure the accuracy of spatial overlay analysis.

[0024] Step 102: Based on the hydrodynamic model, perform evolution simulation on multi-source basic geographic data to generate flood spatiotemporal evolution data, which includes the time-varying inundation depth sequence and water flow velocity sequence.

[0025] This embodiment aims to reconstruct the physical evolution of floods. Specifically, the hydrodynamic model can employ a one-dimensional coupled numerical simulation method. For example, for major rivers and drainage networks, the one-dimensional Saint-Venant equations can be used to simulate the rapid propagation characteristics of water flow along specific cross-sections; while for flat floodplains or urban blocks, the two-dimensional Shallow Water Equations are used for gridded simulation to reflect the spread and storage process of floods on a plane. By solving the model, a four-dimensional spatiotemporal dataset (x,y,t,features) can be output, where x and y represent spatial coordinates, t represents the simulation time step (e.g., one output frame every 10 minutes), and features contains the inundation depth h(x,y,t) and flow velocity v(x,y,t) at that location at that moment. This spatiotemporal evolution data constitutes the basic physical field for all subsequent risk assessments, reflecting not only where the flooding occurs but also when and how rapid the flow is.

[0026] Step 103: Based on the spatiotemporal evolution data of floods and population distribution data, analyze the interaction characteristics between flood intensity and the vulnerability of disaster-bearing bodies, and calculate the personnel safety risk index; Step 104: Based on the spatiotemporal evolution data of floods and transportation network data, analyze the attenuation characteristics of the physical state of transportation facilities and network capacity, and calculate the transportation operation risk index; Step 105: Based on the spatiotemporal evolution data of floods and economic asset data, analyze the exposure degree and value loss rate of different asset types, and calculate the economic loss risk index.

[0027] The three steps described above constitute the three independent pillars of the multidimensional assessment. The personnel safety risk index is a calculated indicator representing the likelihood of injury or entrapment of personnel; a higher value indicates a greater threat to life safety. The interaction characteristic can be understood as the spatial and temporal collision between the disaster-causing factor (flood intensity) and the disaster-bearing body (population). The traffic operation risk index is used to calculate the degree of loss of road network function, considering not only whether roads are flooded but also the impact of water flow speed on vehicle stability and the damage to overall network connectivity caused by local blockages. The economic loss risk index focuses on the reduction in asset value, typically expressed as the ratio of the expected economic loss to the total asset value of the region. These three indices can be normalized during calculation (e.g., mapped to the 0-1 range) to facilitate subsequent comprehensive evaluation. For the specific calculation method of each dimension, either the static overlay method or the dynamic racing method can be selected based on the data conditions.

[0028] Step 106: Based on the personnel safety risk index, traffic operation risk index and economic loss risk index, a multi-dimensional integrated evaluation is conducted to obtain the multi-dimensional comprehensive flood risk assessment results for the target area.

[0029] This embodiment aims to integrate multi-dimensional risk indicators into a unified decision-making view. In some preferred embodiments, Dempster-Shafer Theory can be used for multi-source information fusion. Specifically, the risk indices of personnel, transportation, and economy can be regarded as three independent evidence bodies, and basic probability assignment functions (BPA) can be constructed for each. For example, if a region has extremely high personnel risk but low economic risk, the Dempster synthesis rule m(A)=[Σm1(B)m2(C)] / (1-K) can be used for fusion calculation, which can effectively handle conflicting information and obtain a comprehensive confidence level that takes into account the influence of all aspects. In addition, as an alternative, fuzzy comprehensive evaluation method or analytic hierarchy process (AHP) can be used to determine the weight vector of each dimension through expert scoring (e.g., personnel safety weight 0.5, transportation weight 0.3, economic weight 0.2), and then perform weighted summation to obtain the comprehensive risk level. The results of a comprehensive risk assessment are usually presented in the form of a risk zoning map, which uses different colors (such as red, orange, yellow, and green) to visually display the extremely high-risk areas, high-risk areas, medium-risk areas, and low-risk areas within the region.

[0030] Example 2 describes how to use static features for risk assessment in scenarios with limited data or low timeliness requirements.

[0031] Step 201: Extract the maximum inundation depth distribution from the spatiotemporal evolution data of the flood; Step 202: Spatially overlay the population distribution data with the maximum inundation depth distribution to determine the number of statically exposed people in the flooded area; Step 203: Obtain the population vulnerability factor in the flooded area, which includes at least age structure characteristics and health status characteristics; Step 204: Calculate the personnel safety risk index based on the number of statically exposed people, the maximum inundation depth distribution, and the population vulnerability factor.

[0032] In this embodiment, the personnel risk assessment uses the classic maximum summation logic. Specifically, for each grid cell, the change in water depth over time is no longer considered; instead, the maximum inundation depth H during the entire flood process is directly extracted. max(x, y). By spatially overlaying the maximum water depth layer with the population density layer, the number of permanent residents within the coverage area of ​​different water depth levels (e.g., 0.5 meters to 1.0 meters) can be calculated, i.e., the number of statically exposed populations. Furthermore, to reflect the differences in flood resistance among different population groups, a population vulnerability factor is introduced. For example, this factor can be set to be proportional to the proportion of elderly (over 65 years old) and children (under 14 years old), or it can be correlated with the proportion of disabled persons or patients with chronic diseases in the area. When calculating the personnel safety risk index, a product form can be used, i.e., R = E × V × H; where E is the number of exposed populations, V is the normalized vulnerability factor, and H is the hazard coefficient based on the maximum water depth. As a preferred refinement scheme, to compensate for the shortcomings of static data, a day-night dual-mode assessment can also be introduced. Specifically, two sets of population distribution layers can be preset: one is a daytime mode, where the population is mainly concentrated in business districts (CBD), schools, and industrial areas; the other is a nighttime mode, where the population returns to residential areas. Based on the specific time period (daytime or nighttime) of the flood, the corresponding population layer is selected for overlay calculation to improve the accuracy of static assessment.

[0033] Step 205: Construct a traffic network topology model based on traffic network data to identify key nodes and key road sections in the network; Step 206: Obtain the pre-constructed inundation depth-loss rate curve for traffic facilities; Step 207: Extract the maximum inundation depth at key nodes and key road sections from the spatiotemporal evolution data of the flood; Step 208: Calculate the direct physical loss rate of key nodes and key road sections using the inundation depth-loss rate curve; Step 209: Calculate the network connectivity degradation index based on the direct physical loss rate and the traffic network topology model to obtain the traffic operation risk index.

[0034] For traffic risks, this embodiment focuses on the physical damage to facilities and structural connectivity failures. Road intersections are abstracted as nodes, and road segments as edges, constructing a traffic network topology model G=(V,E). By calculating indices such as betweenness centrality or degree centrality, critical nodes (such as cross-river bridges and interchanges) and critical road segments with the greatest impact on the overall network function are identified. A pre-set inundation depth-loss rate curve is used to calculate the degree of water damage. This curve is typically fitted from historical disaster data; for example, when the water depth is less than 0.3 meters, the roadbed loss rate is 0; when the water depth is between 0.3 meters and 1.0 meter, the loss rate increases linearly with water depth; when the water depth exceeds 2.0 meters, it is considered washed away, with a loss rate of 100%. Substituting the simulated maximum water depth into this curve yields the direct physical loss rate. Assuming that severely damaged road sections (such as those with a loss rate greater than 80%) are removed from the topology graph, the maximum connected subgraph size or average shortest path length of the network is recalculated, and the rate of change of these topology indicators is used as the traffic operation risk index.

[0035] Step 210: Divide the economic asset data into multi-level economic exposure units and determine the asset value of each economic exposure unit; Step 211: Construct a multi-factor loss function, the input variables of which include at least inundation depth, water flow velocity, and inundation duration; Step 212: Extract the inundation depth, water flow velocity, and inundation duration characteristics of the flood spatiotemporal evolution data at each economic exposure unit; Step 213: Calculate the expected economic loss amount for each economic exposure unit using the multi-factor loss function and asset value; Step 214: Obtain the economic loss risk index by normalizing the expected economic loss amount.

[0036] Specifically, the calculation formula is: DL facility =SUM(AV i *LR i ); where DL facility This represents the total direct economic loss to transportation facilities; SUM is the summation operation, which sums up all damaged facilities within the disaster area; i is the number of the transportation facility; AV i Let LR be the asset value of the i-th transportation facility; i Let be the loss rate for the i-th transportation facility, determined based on the flooding depth.

[0037] Loss total =Loss direct +SUM(IF ij *Loss indirect_j ); where Loss total Total loss for the economically exposed unit; Loss directThis represents the direct economic loss calculated based on a multi-factor loss function; SUM represents the summation operation; IF represents the direct economic loss. ij The impact factor of associated node j on current node i, derived from the input-output table or supply chain matrix; Loss indirect_j The amount of indirect loss passed to the associated node j.

[0038] In this embodiment, economic risk assessment is broken down into two parts: direct losses and indirect losses. The study area is divided into economic exposure units at different levels, such as industrial parks, commercial districts, or specific individual buildings, and their asset value (AV) is estimated based on land use type and building area. Instead of relying solely on water depth, a multi-factor loss function DL=f(h,v,t)×AV is constructed; where h is the flooded water depth, v is the flow velocity (high flow velocity may lead to building collapse), and t is the flooding duration (long soaking time may lead to equipment damage). Substituting these physical characteristics into the function, direct economic losses can be calculated. Based on this, as a preferred implementation, this embodiment can also introduce a cascade failure model of the economic network to assess indirect losses. Specifically, a supply chain network between enterprises is constructed based on the input-output table. When a node (such as a key component factory) is damaged by flooding and stops production, this functional failure will propagate along the supply chain to downstream nodes, causing a decline in output value for enterprises in unflooded areas as well. By summing the direct losses and the indirect losses generated by the cascade, and then normalizing them, we can obtain an economic loss risk index that comprehensively reflects the resilience of the regional economy.

[0039] Example 3 describes how to construct a time-dependent traffic network (TDG) using continuous flood simulation data. Unlike static topology graphs, the network constructed in this example can reflect the dynamic decay process of road capacity during flood spread, providing an accurate time-varying road network foundation for subsequent dynamic evacuation calculations.

[0040] Step 301: Divide the target area into grid cells according to the preset spatial resolution; for each grid cell in the target area, extract the moment when the flood depth sequence or water flow velocity sequence first exceeds the preset personnel safety threshold in the flood spatiotemporal evolution data to obtain the personnel danger arrival time data.

[0041] In this embodiment, the extraction of personnel arrival time data is a process of temporal scanning of flood spatiotemporal evolution data. Specifically, for each spatial grid cell g, the system iterates through its water depth sequence h(g,t) and current velocity sequence v(g,t) within the simulation time period T. Preset personnel safety thresholds can be set based on human instability mechanisms, for example, a water depth threshold of 0.5 meters, or a threshold for the product of water depth and current velocity of 1.5 square meters per second. When the values ​​in the sequence first meet any of the above conditions, that moment is marked as the personnel arrival time T for that grid cell. risk (g) If a grid does not exceed the threshold throughout the entire simulation, then its T... risk Marked as infinity or the end of the simulation. This data is stored in the form of a raster matrix, visually representing the spatial arrival wavefront of the flood threat.

[0042] T risk (g)=min{t|h(g,t)≥h limit Or v(g,t)≥v limit}; where T risk (g) represents the arrival time of personnel in grid cell g; min{...} represents the earliest time that satisfies the conditions; t is the simulation time step; h(g,t) is the submerged water depth of grid cell g at time t; v(g,t) is the water flow velocity of grid cell g at time t; h limit The preset safe water depth threshold for personnel; v limit This is the preset threshold for safe personnel flow rate.

[0043] Step 302: For each road segment in the traffic network data, extract the moment when the flood depth sequence or water flow velocity sequence first exceeds the preset vehicle passage threshold in the flood spatiotemporal evolution data to obtain the road segment blockage arrival time data.

[0044] This embodiment focuses on the linear element of a road segment. Since vehicles typically have a greater wading capability than pedestrians, the preset vehicle passage threshold is usually set higher. For example, for sedans, the water depth threshold can be set to 0.3 meters (exhaust pipe height); for large emergency vehicles, this threshold can be set to 0.6 meters or higher. For each road segment e, the system obtains hydrodynamic data from all grid cells covering the segment through spatial mapping, typically taking the maximum water depth or maximum flow velocity in these grid cells as the state value of the segment at time t. A time-series scan is performed to find the moment when the vehicle passage threshold is first exceeded, and this moment is recorded as the road segment closure arrival time T. block (e).

[0045] Step 303: Construct a time-dependent transportation network. The edge attributes of the time-dependent transportation network include the traffic capacity and travel time that change over time.

[0046] In this embodiment, the time-dependent traffic network is formally defined as a dynamic graph G(t) = (V, E, C(t), τ(t)); where V represents the set of nodes and E represents the set of edges. Unlike traditional static graphs, each edge e in the graph no longer has a fixed weight, but is assigned two functional properties that change with time t: dynamic capacity C. e (t) and dynamic passage time τ e (t). In terms of data storage structure, it can be stored using a sequence of discrete time-slice snapshots, such as storing a network state graph every 5 minutes; or it can be stored using piecewise linear functions or continuous function objects to store edge attributes, supporting queries at any consecutive time.

[0047] Step 304: For each road segment in the time-dependent traffic network, based on the flood spatiotemporal evolution data of the flood inundation depth sequence and water flow velocity sequence at that road segment, the dynamic traffic capacity of that road segment at different time steps is calculated using a preset capacity decay function.

[0048] This embodiment represents a crucial step in mapping hydrodynamic physical fields to traffic network performance. The system reads the inundation depth sequence h corresponding to each road segment. e (t) and water flow velocity sequence v e The remaining capacity of the road segment at time t is calculated by substituting the variable (t) into a preset capacity decay function. The output of this function is the remaining capacity of the road segment at time t. This mechanism ensures that the state of the traffic network is driven by the flood evolution process, accurately reflecting the physical phenomena of rising water levels and road congestion.

[0049] Step 305, the preset expression for the capacity decay function is: C e (t)=C e0 ·exp(-a h ·(h e (t) / h0) p -a v ·(v e (t) / v0) q ); where C e (t) represents the dynamic traffic capacity of the road segment at time t; C e0 The initial design capacity of the road segment is represented by exp(·); exp(·) represents an exponential function with the natural constant e as its base; h e (t) represents the submerged water depth of the road section at time t; v e (t) represents the water flow velocity of the road segment at time t; h0 represents the preset water depth characteristic reference value; v0 represents the preset flow velocity characteristic reference value; a h This indicates the preset water depth influence weighting coefficient; a vThis represents the preset weighting coefficient for the influence of flow velocity; p and q represent preset nonlinear exponential parameters.

[0050] In this embodiment, an exponential decay function is used to simulate the psychological and physical deceleration effects on a driver when facing standing water. Specifically, parameters h0 and v0 serve a normalization function; for example, h0 is set to 0.3 meters. When the actual water depth h... e When (t) is small, the ratio is close to 0, the exponential term is close to 1, and the traffic capacity is almost unaffected; as the water depth increases, the exponential term decreases sharply, leading to a rapid decline in traffic capacity. Parameters p and q control the steepness of the decay curve; for example, taking p=2 indicates that the effect of water depth exhibits a non-linear and rapidly deteriorating characteristic. A specific numerical calculation example: Assume the initial traffic capacity C of a certain road segment... e0 For a speed of 2000 vehicles per hour, parameter a h =1, h0=0.3 meters, p=2 (ignoring the influence of flow velocity). When the flood evolves to a certain moment t1, and the water depth in this section reaches 0.15 meters, the calculated exponential term is exp(-(0.15 / 0.3)). 2 When the water depth reaches 0.3 meters, the exponential term becomes exp(-0.25)≈0.778, at which point the traffic capacity drops to approximately 1556 vehicles per hour; when the water depth reaches 0.3 meters, the exponential term becomes exp(-1)≈0.367, and the traffic capacity drops sharply to approximately 734 vehicles per hour. This continuous function expression more accurately reflects the gradual changes in traffic flow than simple on / off binary logic.

[0051] Step 306: Based on the road blockage arrival time data, set the dynamic traffic capacity value after the blockage arrival time to zero or a preset minimum value close to zero.

[0052] This embodiment represents a hard constraint truncation of the continuous decay process. Although gradual decay is calculated, in real-world physical scenarios, when the water depth exceeds the vehicle's wading limit (i.e., T...),... block After a certain point (t>T), the road is effectively paralyzed. Therefore, the system will force a change in the timeframe from t>T. block (e) C during the time period e (t) is set to 0, and the corresponding travel time τ is also set to 0. e (t) is set to infinity. This ensures that the algorithm will not select these physically interrupted road segments in the subsequent shortest path search, thus guaranteeing the safety and feasibility of the evacuation route.

[0053] Step 307: Calculate the dynamic service level of the road network based on the time-dependent traffic network to obtain the traffic operation risk index.

[0054] After constructing the aforementioned time-dependent transportation network, the system can assess traffic risk by calculating dynamic evolution indicators of network flows. For example, it can calculate the proportion of total network capacity (Capacity-Hours) lost during the entire flood evolution process relative to the state without flooding, or calculate the dynamic growth rate of the network's average travel time. After normalizing these macro-indicators, a traffic operation risk index can be obtained.

[0055] Example 4 describes a dynamic risk assessment method for personnel based on evacuation feasibility. Compared with assessment methods that only consider static exposure, this example introduces the concept of a race in the time dimension, that is, comparing the difference between the time of flood arrival and the time of personnel evacuation, which can distinguish between two fundamentally different risk scenarios: deep water areas where people can escape and shallow water areas where people cannot escape.

[0056] Step 401: For each assessment unit within the target area, obtain the arrival time data of personnel hazards.

[0057] In this embodiment, the evaluation unit can be a regular raster grid or a community patch based on census data. For each evaluation unit i, the system directly calls T. risk (i) Data. If the evaluation cell contains multiple grids, in order to follow the worst-case scenario principle, the earliest hazard arrival time among all grids in the area is usually taken as the T value for that cell. risk (i) means that the area is considered to be in a state of risk as soon as any point in the area begins to become dangerous.

[0058] Step 402: Based on the time-dependent transportation network, calculate the shortest evacuation time from the assessment unit to the nearest refuge site.

[0059] Unlike calculating the shortest path on a static graph, this embodiment requires pathfinding on a time-dependent traffic network G(t). Specifically, it employs a time-dependent Dijkstra's algorithm or the A* algorithm. When searching for a path, the algorithm considers the time t required to reach a given edge e. entry Query the dynamic travel time τ of this edge at this moment. e (t entry If you depart late, some sections of the road may become congested due to flooding (τ). e (increase) or even interrupt (τ) e (If the value is infinite, vehicles must detour, increasing evacuation time.) The system iterates through all candidate refuge locations, calculates the time taken to reach each refuge point from evaluation unit i, and takes the minimum value as the shortest evacuation time T. travel (i,t). It should be noted that t here refers to the departure time, which is usually the current assessment time or the time when the warning is issued.

[0060] Step 403: Calculate the evacuation time margin, which represents the time difference between the arrival time of personnel in danger and the time required to complete the evacuation.

[0061] Evacuation time margin M is a direct physical quantity that measures the level of safety. Its calculation logic is: when will the flood arrive (T)? risk Subtract the current moment (t), subtract the time it takes for people to react and prepare to leave (response delay), and then subtract the time spent running on the road (T). travel If the result is positive, it means that the person ran faster than the flood and had a safety margin; if the result is negative, it means that the person had not reached the refuge point before the flood caught up with them, i.e., they were trapped.

[0062] Step 404, the formula for calculating the evacuation time margin and the probability of successful evacuation is: M(i,t)=T risk (i)-t-ΔT resp -T travel (i,t); P evac (i,t)=1 / (1+exp(-k·(M(i,t)-b))); where M(i,t) represents the evacuation time margin of the i-th evaluation unit at time t; T risk (i) represents the time of arrival of personnel at risk in the assessment unit; ΔT resp Indicates the preset personnel response delay time; T travel (i,t) represents the shortest evacuation time calculated based on the time-dependent transportation network at time t; P evac (i,t) represents the probability of successful evacuation of the i-th evaluation unit at time t; k and b represent the preset probability mapping parameters.

[0063] This embodiment maps the physical time difference to a probability value, reflecting the uncertainty in reality. The formula uses ΔT... respThis is a key parameter that can be set according to population characteristics. For example, for an elderly community, the response delay may be as long as 30 minutes; while for a young and middle-aged community, it may only take 10 minutes. The probability mapping function adopts the form of a Sigmoid function. Here, parameter b represents the critical time margin. For example, if b=10 minutes, even with a theoretically calculated margin of 10 minutes, the probability of successful evacuation is only 50%, which reserves a safety margin for dealing with emergencies (such as vehicle breakdowns, panic, and congestion). Parameter k controls the sensitivity to probability changes; the larger the k value, the steeper the curve, indicating greater sensitivity to changes in the time margin. For example: Suppose a flood will arrive in a certain area in 60 minutes, the current time is 0, the response delay is 10 minutes, and the calculated evacuation time is 30 minutes. Then M=60-0-10-30=20 minutes. If b=10, k=0.2, then Mb=10, exp(-2)≈0.135, P evac =1 / (1+0.135)≈0.88, meaning there is an 88% probability of a successful evacuation. Conversely, if traffic congestion increases the travel time to 45 minutes, then M=5 minutes, Mb=-5, exp(1)≈2.718, P evac =1 / (1+2.718)≈0.27, the evacuation success rate plummeted to 27%.

[0064] Step 405: Calculate the probability of successful evacuation based on the evacuation time margin.

[0065] This embodiment achieves a closed-loop risk assessment model. The system uses the calculated probability of successful evacuation to reduce or correct traditional static risks.

[0066] Step 406: Based on the probability of successful evacuation, the static flood risk is dynamically corrected to obtain the personnel safety risk index, which is calculated using the following formula: R(i,t)=H(i,t)·V(i)·(1-P evac (i,t)); where R(i,t) represents the personnel safety risk index of the i-th assessment unit at time t; H(i,t) represents the flood hazard factor of the assessment unit at time t, the value of which is determined based on the inundation depth and flow velocity in the flood spatiotemporal evolution data; V(i) represents the population vulnerability factor of the assessment unit; P evac (i,t) represents the probability of successful evacuation of the evaluation unit at time t.

[0067] This formula indicates that risk depends not only on the depth of the water (H) and the vulnerability of the person (V), but also on whether the person can escape (1-P). evac When P evac When P approaches 1 (i.e., evacuation is extremely easy), regardless of the water depth, the adjusted risk R will approach 0, which aligns with the logic of an empty city being free from disaster; when P... evacWhen the water level approaches zero (i.e., evacuation is impossible), the risk R returns to a high value determined by the hazard H and vulnerability V. This dynamic assessment model can accurately identify those seemingly shallow water areas that are isolated due to traffic disruptions, or areas that are deep but have low actual risk of casualties due to good evacuation conditions, providing a highly scientific basis for the precise deployment of emergency rescue forces.

[0068] Regarding boundary condition handling, when the calculated shortest evacuation time T... travel When (i,t) is infinite (i.e., all paths originating from this assessment unit have been blocked by floodwaters), the system's probability of successfully evacuating this assessment unit is P. evac (i,t) is set to zero, and the area is marked as an isolated area in the output. As the highest priority target for emergency rescue, the isolated area will automatically calculate its straight-line distance to the nearest passable area, supporting the dispatching decisions of rescue ships or helicopters.

[0069] Based on the above embodiments, another implementation method is provided: when high-precision road network data is unavailable or computing power is limited, the continuous probability correction can be simplified into a hierarchical correction matrix. For example, the area can be divided into three levels—allowable evacuation, urgent evacuation, and difficult evacuation—based on evacuation time margin, corresponding to risk correction coefficients of 0.2, 0.8, and 1.0, respectively, to achieve a lightweight dynamic assessment.

[0070] Example 5 describes a preferred high-precision implementation method for acquiring population distribution data. Traditional flood risk assessments often directly use static population statistics from administrative divisions, which are insufficient to reflect the true population distribution during disasters (e.g., weekday daytime or holiday nighttime). While LBS (Location-Based Services) data can provide real-time distribution, it is often subject to systematic biases due to limitations in base station coverage and user device usage. This example addresses this problem by introducing confidence and correction mechanisms.

[0071] Step 501: Obtain static baseline population data and real-time LBS activity intensity data for the target area.

[0072] In this embodiment, static base population data typically refers to the gridded resident population data released by the census, which is characterized by accurate total population control but low spatial resolution. Real-time LBS activity intensity data refers to mobile signaling data or location request data from internet applications provided by operators. It should be noted that what operators directly provide is usually the number of active users or location requests within each base station or grid at a specific moment, which is not equivalent to the actual population size, hence the term "activity intensity." For example, in a commercial center, LBS data shows 5,000 active signals, but this may only represent a portion of the actual population in that area.

[0073] Step 502: Calculate the coverage correction coefficient based on the spatial coverage characteristics of the communication base station and the holding rate characteristics of the mobile terminal.

[0074] This embodiment aims to eliminate systematic biases. The coverage correction coefficient is a multiplier used to restore the signal strength to the number of people. Specifically, the spatial coverage characteristics of communication base stations reflect signal blind spots. For example, in mountainous areas or underground spaces, base station coverage may be only 80%. Mobile terminal ownership characteristics reflect the proportion of people who own mobile phones. For example, in elderly communities or primary schools, ownership may be only 60%; while in business districts, due to the phenomenon of one person owning multiple phones, ownership may exceed 100%. The system calculates the correction coefficient α(g) for each grid cell g based on the base station distribution layer and land use type (inferring population structure) in the Geographic Information System (GIS). For example, if the base station coverage of a certain grid is R... cov =0.9, average holding rate R phone =0.8, then the correction coefficient α(g) = 1 / (R cov ×R phone )≈1.39.

[0075] Step 503: Correct the real-time LBS activity intensity data using the coverage correction coefficient to obtain dynamic population increment data.

[0076] Based on the above calculations, the system multiplies the original LBS activity intensity L(g,t) by the correction coefficient α(g) to obtain the corrected dynamic population estimate, i.e., the dynamic population increment data. This process maps the digital signal space back to the physical population space.

[0077] Step 504: Calculate the data confidence index based on the time variation coefficient and spatial variation coefficient of real-time LBS activity intensity data.

[0078] Despite corrections, LBS data may still exhibit abnormal fluctuations due to signal drift or equipment malfunction. This embodiment uses statistical indicators to assess the reliability of the data. Specifically, the coefficient of variation (CV) is used to construct a data confidence index. (CV is a time-varying coefficient of variation.) t This reflects the numerical stability of the same location within a short time window. If the value fluctuates wildly (e.g., 100 people one minute, 0 people the next), the data is unreliable; the coefficient of spatial variation (CV) sThis reflects the smoothness of adjacent grids. The data confidence index Conf(g,t) can be constructed as the inverse function of CV. For example, the formula Conf(g,t) = 1 - min(1, CV(g,t) / CV0) can be used; where Conf(g,t) is the data confidence index of grid g at time t; min is the minimum value function; CV(g,t) is the time variation coefficient of real-time LBS data; and CV0 is the preset tolerance threshold for the variation coefficient. When the data fluctuates greatly, the confidence index approaches 0; when the data is stable, the confidence index approaches 1.

[0079] Step 505: The static baseline population data and the dynamic population increment data are weighted and fused, and the data confidence index is associated to generate dynamic population distribution data, which is then used as population distribution data for subsequent calculations.

[0080] This embodiment achieves intelligent fusion of multi-source data. The fusion formula can be expressed as: Pop final (g,t)=(1-Conf(g,t))×Pop static (g)+Conf(g,t)×Pop dynamic (g,t); where Pop final (g,t) represents the generated dynamic population distribution data; Pop static (g) represents static baseline population data; Pop dynamic (g,t) represents the LBS dynamic population increment data after coverage correction; Conf(g,t) is the data confidence index. The physical meaning of this formula is: when the confidence level of the LBS data is high, the system primarily adopts real-time dynamic population data to reflect the actual flow of people; when the confidence level of the LBS data decreases due to base station failure or signal interference, the system automatically degrades and relies more on static baseline population data to ensure the robustness of the assessment system and avoid significant deviations in risk assessment results due to missing data sources.

[0081] Example 6: This example not only assesses risk but also supports decision-making. In situations with limited emergency rescue resources, commanders often face the difficult decision of which road to clear first. This example calculates the net value of each road for saving lives by introducing the concept of marginal contribution from economics and the counterfactual reasoning approach from causal inference.

[0082] Step 601: Select candidate road segments for emergency access from the time-dependent traffic network.

[0083] Candidate road sections for emergency repair refer to roads that are currently blocked or about to be blocked by floods, but have the potential for emergency repair. The system can determine this set by setting filtering rules. For example, filtering rules may include: road classification as arterial road or evacuation route; and the current time T of blockage arrival for the road section.block The road segment is either less than the latest evacuation time for personnel, or it is located on a major thoroughfare in a densely populated area. These road segments constitute the solution space for decision-making.

[0084] Step 602: Based on the evacuation success probability and population distribution data in the personnel safety risk index, calculate the marginal contribution of each candidate road segment for emergency clearance; the marginal contribution represents the expected reduction in the number of trapped people in the target area under the assumption that the candidate road segment for emergency clearance remains passable or has improved passability within a specific time window.

[0085] The marginal contribution ΔN defined here is a difference index. To calculate it, the system needs to perform one baseline calculation and one intervention calculation.

[0086] ΔN(e)=N trapped_base -N trapped_intervention Where ΔN(e) is the marginal contribution of candidate road segment e; N trapped_base N represents the baseline number of people trapped in the no-intervention scenario; trapped_intervention The number of people trapped after intervention to clear or postpone the blocked road section e.

[0087] Step 603: Based on the current time-dependent transportation network, calculate the baseline evacuation success probability of the target area under the no-intervention scenario, and calculate the number of people trapped at the baseline by combining population distribution data.

[0088] This is the control group for the simulation. The system calculates the baseline evacuation success probability P for each assessment unit i on a current time-dependent traffic network without any human intervention. base (i). Subsequently, combining the population distribution data Pop(i), the total baseline number of trapped people N is calculated. trapped_base =Σ[Pop(i)×(1-P base (i))]. This value represents how many people are expected to be trapped if no action is taken.

[0089] ΔN win (e)=SUM t (w(t)*(N trapped_base (t)-N trapped_intervention (t)));where ΔN win (e) represents the cumulative marginal contribution of candidate road segment e over the entire evaluation time window; SUM t This is a summation over time step t; w(t) is the time weighting coefficient (e.g., the closer to the flood peak, the greater the weight); N trapped_base (t) and N trapped_intervention (t) represents the baseline at time t and the number of people trapped by the intervention, respectively.

[0090] Step 604: Construct an intervention scenario, modify the dynamic traffic capacity of candidate road segments in the time-dependent traffic network to a mandatory traffic state, or postpone the arrival time of their road segment blockage by a preset time.

[0091] This is the experimental group for the simulation. The system creates a temporary copy of the time-dependent traffic network in memory and modifies the parameters for a specific candidate road segment e. This embodiment provides two specific modification methods, corresponding to different engineering approaches: the first is a forced passage state, that is, the traffic capacity C of the road segment during floods. e (t) Forced locking to a non-zero value physically corresponds to dispatching an amphibious rescue vehicle or erecting an elevated temporary bridge; ΔN win (e)=SUM t (w(t)*(N trapped_base (t)-N trapped_intervention (t)));where ΔN win (e) represents the cumulative marginal contribution of candidate road segment e over the entire evaluation time window; SUM t This is a summation over time step t; w(t) is the time weighting coefficient (e.g., the closer to the flood peak, the greater the weight); N trapped_base (t) and N trapped_intervention (t) represents the baseline at time t and the number of people trapped by the intervention, respectively. The second method is to postpone the blocking time, that is, to delay the blocking time of T for this section. block (e) Increase Δt (e.g., 2 hours), which physically corresponds to the time it takes to delay flooding by piling up sandbag walls along the roadside. T block_interv (e)=T block (e)+ ΔT delay Among them, T block_interv (e) represents the arrival time of the blockage of road segment e under the intervention scenario; T block (e) represents the moment when the floodwaters block the circuit under the baseline scenario; ΔT delay This represents the time delay (e.g., 2 hours) gained by engineering measures. This formula is used to simulate the effect of sandbag dikes or seepage prevention projects in delaying road flooding.

[0092] Step 605: Based on the modified time-dependent transportation network, recalculate the probability of successful intervention and evacuation in the target area and the number of trapped people under the intervention scenario.

[0093] Using the modified network replica, the system runs the shortest path search and probability calculation algorithm again. Because the traffic conditions on road segment e have improved, the shortest evacuation time T for some evaluation units i that previously required detours or had no way to travel has decreased. travel (i) This shortens the time, increasing the probability of a successful evacuation, P. intervention (i). The system calculates the number N of the trapped population after the intervention. trapped_intervention .

[0094] Step 606: Calculate the difference between the baseline number of trapped people and the number of trapped people receiving intervention to obtain the marginal contribution of the candidate road segment for clearing the road.

[0095] The system performs a subtraction operation: ΔN(e) = N trapped_base -N trapped_intervention This difference ΔN(e) intuitively calculates how many lives were saved by clearing road segment e. The system repeats the above process for all candidate road segments, sorting them from largest to smallest by marginal contribution to generate a priority list for disaster relief. As an example, assuming the marginal contribution of road segment A is 500 lives and the marginal contribution of road segment B is 50 lives, even if the repair cost of road segment B is lower, the system will still recommend prioritizing the repair of road segment A because its life-saving value is higher.

[0096] Example 7 describes the source of the probability mapping function parameters (k and b). These parameters determine the accuracy of the risk assessment model, and their acquisition process reflects the data-driven scientific nature of this invention.

[0097] Step 701, the calculation of the personnel safety risk index involves using a probability mapping function to map the evacuation time margin to the probability of successful evacuation; before calculating the personnel safety risk index, the parameters of the probability mapping function are pre-calibrated.

[0098] This example illustrates that this is an offline model training or parameter calibration process, typically performed before system deployment or during annual updates, rather than every time during real-time evaluation.

[0099] Step 702: Collect data on personnel evacuation cases in historical flood disasters. The personnel evacuation case data includes historical evacuation time margins and corresponding evacuation result labels.

[0100] To obtain accurate parameters, historical data needs to be reviewed. Researchers collected data on real flood disasters from the past and extracted several individual evacuation cases through retrospective analysis. For each case, the historical evacuation time margin (independent variable x) was calculated by simulating the water and traffic conditions at the time. At the same time, based on post-disaster investigation records, the final status of the individual was determined (successful evacuation was recorded as 1, and being trapped or injured was recorded as 0, i.e., dependent variable y).

[0101] Step 703: Construct a logistic regression model with evacuation time margin as the independent variable and evacuation outcome label as the dependent variable.

[0102] Logistic Regression is a classic method for handling binary classification problems and outputting probability values. The model constructed in this embodiment is in the form: P(y=1|x)=1 / (1+exp(-(β1x+β0))); where P(y=1|x) is the probability of successful evacuation (y=1) given an evacuation time margin x; β1 is the regression coefficient of the logistic regression model; and β0 is the intercept term. Comparing this to the Sigmoid formula 1 / (1+exp(-k(Mb))), it can be seen that the two are mathematically equivalent, where k corresponds to β1 and k×b corresponds to -β0.

[0103] Step 704: Use the personnel evacuation case data to train the logistic regression model using maximum likelihood estimation, calculate the shape parameters and position parameters of the probability mapping function, and store them as preset parameters.

[0104] The collected case data is input into the model, and the Maximum Likelihood Estimation (MLE) method is used iteratively to maximize the joint probability of the observed data, yielding the optimal regression coefficients β1 and β0. These coefficients are then converted into a shape parameter k (reflecting the steepness of the probability change with the margin) and a location parameter b (reflecting the margin threshold when the probability is 0.5). These parameters are stored in the system's configuration file or database for use by the real-time risk assessment module. In some optional implementations, multiple sets of parameters can be trained and stored separately for different age groups or different terrain regions (e.g., mountainous and plain areas) to achieve more refined assessments.

[0105] In a specific calibration case, data on 200 evacuation cases in a city over the past 10 years were collected, including 150 successful evacuations and 50 cases of people being trapped. Evacuation time margin was used as the independent variable, and the evacuation outcome (success / trapped) as the dependent variable, inputting the data into a logistic regression model. Through maximum likelihood estimation training, parameters k=0.23 and b=8.5 minutes were obtained. This parameter combination indicates that: when the evacuation time margin is 8.5 minutes, the probability of successful evacuation is 50%; when the margin increases to 20 minutes, the probability of success rises to approximately 94%; when the margin is negative (i.e., theoretically impossible to evacuate), the probability of success drops to below approximately 10%. This parameter combination achieved a prediction accuracy of 82% in five-fold cross-validation.

[0106] Example 8 describes the system architecture and hardware entity for implementing the above method.

[0107] Step 801, Data Acquisition Module, is used to acquire multi-source basic geographic data of the target area. The multi-source basic geographic data includes at least hydrological and meteorological data, topographic and geomorphological data, population distribution data, transportation network data, and economic asset data.

[0108] The data acquisition module can be configured as a multi-channel data interface server. Specifically, it includes a network interface for connecting to the meteorological bureau's API, an I / O interface for reading local GIS databases (such as PostGIS), and a message queue consumer (such as a Kafka client) for receiving real-time LBS data streams.

[0109] Step 802, the simulation module, is used to perform evolution simulation on multi-source basic geographic data based on the hydrodynamic model to generate flood spatiotemporal evolution data, which includes the time-varying inundation depth sequence and water flow velocity sequence.

[0110] The simulation module is typically deployed on servers equipped with high-performance graphics processing units (GPUs) or computing clusters. It integrates pre-compiled hydrodynamic solvers (such as CUDA-accelerated solvers) and is capable of rapidly parallelizing water flow evolution calculations involving millions of grid cells.

[0111] Step 803, the personnel risk assessment module, is used to analyze the interaction characteristics between flood intensity and the vulnerability of disaster-bearing bodies based on flood spatiotemporal evolution data and population distribution data, and calculate the personnel safety risk index; Step 804, the traffic risk assessment module, is used to analyze the attenuation characteristics of the physical state of traffic facilities and network capacity based on flood spatiotemporal evolution data and traffic network data, and calculate the traffic operation risk index; Step 805, the economic risk assessment module, is used to analyze the exposure degree and value loss rate of different asset types based on flood spatiotemporal evolution data and economic asset data, and calculate the economic loss risk index.

[0112] These three assessment modules constitute the system's application logic layer, which can be implemented as independent microservice containers. The personnel risk assessment module encapsulates the shortest path search algorithm and probability mapping logic; the traffic risk assessment module maintains time-dependent traffic network objects in memory; and the economic risk assessment module integrates an asset value database and a multi-factor loss function library.

[0113] Step 806, the comprehensive assessment module, is used to conduct a multi-dimensional fusion evaluation based on the personnel safety risk index, traffic operation risk index and economic loss risk index to obtain the multi-dimensional comprehensive flood risk assessment results for the target area.

[0114] The comprehensive evaluation module is responsible for data aggregation and visualization. It includes a weighted fusion engine (executing AHP or DS algorithms) and a graphics rendering engine. The evaluation results can be displayed on a command center screen through a WebGIS interface, allowing users to zoom, query, and perform scenario simulations. In terms of hardware implementation, the system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of the methods described in the above embodiments.

[0115] Example 9 describes how, based on the DS evidence theory, this example further introduces a multi-agent system and a deep learning model to solve the problems of consistency in risk assessment and prediction of future risk trends in complex scenarios.

[0116] Step 106: Based on the personnel safety risk index, traffic operation risk index and economic loss risk index, a multi-dimensional integrated evaluation is conducted to obtain the multi-dimensional comprehensive flood risk assessment results for the target area.

[0117] w i (t,s)=w i0 *α(t)*β(s); where, w i (t,s) represents the dynamic weights of the i-th risk dimension at time t and scenario s; w i0 The base weights are: α(t) is the time decay function (e.g., as the disaster progresses, the risk weight of personnel gradually decreases, while the recovery weight gradually increases); β(s) is the scenario adjustment factor.

[0118] In this embodiment, a multi-agent risk assessment system is constructed to address cognitive uncertainty and ambiguity in risk assessment. Specifically, the system creates independent agents for personnel risk, traffic risk, economic risk, and ecological risk, respectively. Each agent is responsible for the fuzzy evaluation of a specific risk dimension, and the traditional fuzzy comprehensive evaluation is extended using intuitionistic fuzzy set theory during the evaluation process. Intuitionistic fuzzy sets not only consider membership degrees but also introduce non-membership degrees and hesitation degrees, more comprehensively characterizing the hesitation of assessment experts or models when faced with extreme data. After each agent reaches a preliminary conclusion, the system initiates a negotiation mechanism. When the assessment results of different risk dimensions conflict (e.g., extremely high personnel risk but extremely low economic risk), the agents conduct multiple rounds of negotiation through preset weighted average rules and reliability redistribution methods until a consistent evaluation result is reached.

[0119] R(x,y,t+Δt)=R(x,y,t)+SUM ij [D ij *(R j -R i )*Δ t]; where R(x,y,t+Δ t SUM represents the risk value at the next time step (x, y); R(x, y, t) represents the risk value at the current time step; ij For summing over the neighboring grid; D ij R is the spatial diffusion coefficient; j With R i These are the risk values ​​of neighboring grid j and the current grid i, respectively; Δ t For time step.

[0120] Furthermore, to bridge the gap between current situation assessment and trend prediction, this embodiment can also introduce a time series forecasting model. The input to the time series forecasting model is a multidimensional risk index sequence over several past time steps, and the output is a predicted risk change trend for future periods (e.g., the next 1 to 3 hours). In specific implementations, Long Short-Term Memory (LSTM) networks or other machine learning models suitable for time series data can be used. The model's training data comes from simulated data of historical flooding events and is trained through supervised learning. Based on this forecasting model, the system can not only output the current comprehensive risk level but also provide early warnings of risk trends, supporting the proactive deployment of emergency resources.

[0121] In classifying risk levels, this embodiment employs an improved K-means clustering algorithm or DBSCAN density clustering algorithm for automatic classification, avoiding the subjectivity of manually setting thresholds. Simultaneously, a dynamic update mechanism for risk levels is established, automatically adjusting risk level boundaries based on real-time monitoring data and early warning information. For example, when rainfall consistently exceeds historical extremes, the system automatically lowers the trigger threshold for high-risk levels, reflecting the dynamic adaptability of risk assessment.

[0122] Example 10: Supplementary technical details regarding recovery capabilities and visualization. This example emphasizes that risk assessment focuses not only on loss but also on recovery, i.e., the resilience of the system.

[0123] Step 104: Based on the spatiotemporal evolution data of floods and traffic network data, analyze the attenuation characteristics of the physical state of traffic facilities and network capacity, and calculate the traffic operation risk index.

[0124] This embodiment further incorporates the recovery time dimension when calculating traffic risk. Specifically, based on historical disaster data and facility types, a recovery time prediction model is established for different degrees of damage. For example, for roads only submerged in shallow water (water depth less than 0.3 meters), the recovery time depends only on the receding water time; while for roads whose roadbeds have been washed away by deep water, the recovery time must take into account the engineering repair cycle. The model also considers the impact of emergency repair capabilities, resource allocation efficiency, and weather conditions on recovery time, calculates the traffic system function recovery curve, and assesses the evolution of the post-disaster traffic service level gradually recovering over time. The formula for calculating the traffic risk index is revised to: R traffic =α×R loss +β×R interruption +γ×R recovery ;where R recovery The risk of recovery time is represented by weighting coefficients α, β, and γ, which can be configured according to emergency management objectives.

[0125] Step 105: Based on the spatiotemporal evolution data of floods and economic asset data, analyze the exposure degree and value loss rate of different asset types, and calculate the economic loss risk index.

[0126] Resilience Index =(Economy pre -Economy min ) / T recovery Among them, Resilience Index For economic resilience index; Economy pre The pre-disaster economic level; Economy min The minimum economic level during a disaster; T recovery The recovery time required for the economy to return to normal levels.

[0127] Similarly, a recovery cost model based on the "Build Back Better" concept is introduced into economic risk assessment. This model distinguishes between the costs of restorative reconstruction (restoring to the pre-disaster state) and resilience enhancement (strengthening disaster resistance through reinforcement). An economic resilience index is introduced, whose calculation formula can be expressed as: Resilience Index = (Pre-disaster economic level - Minimum economic level) / Recovery time. This index can calculate the economic system's ability to withstand shocks and rebound rapidly after a disaster.

[0128] Step 106: Based on the personnel safety risk index, traffic operation risk index and economic loss risk index, a multi-dimensional integrated evaluation is conducted to obtain the multi-dimensional comprehensive flood risk assessment results for the target area.

[0129] To visually represent the complex, multi-dimensional assessment results, this embodiment developed a three-dimensional risk landscape visualization system. This system utilizes a three-dimensional geographic information engine (such as Cesium or Unreal Engine), with terrain elevation representing risk intensity, to achieve an intuitive display of the spatial distribution of risk (Risk Landscape). Specifically, on the three-dimensional map, areas with higher risk exhibit higher terrain elevations, forming risk peaks. Simultaneously, a spatiotemporal dynamic risk evolution animation is constructed, allowing users to dynamically view the flow and diffusion of risk across time and space using a timeline slider.

[0130] Furthermore, the system integrates virtual reality (VR) and augmented reality (AR) technologies. In emergency drill scenarios, commanders wearing VR headsets can immerse themselves in observing flooded city streets and see risk data labels overlaid on buildings and roads (such as the estimated number of people trapped and road closure countdowns). This immersive interactive experience enhances commanders' understanding of risk scenarios and improves the intuitiveness and accuracy of emergency response decisions. In addition, an interactive risk scenario analysis platform has been established, supporting user-defined parameters (such as adjusting flood control dike height and modifying emergency plans) to simulate and compare the risk evolution results under different intervention measures in real time, enabling "what-if" scenario deduction.

[0131] To address the issue that existing technologies neglect the dynamic interaction between flood evolution and personnel evacuation, thus failing to accurately answer the question of whether escape is possible, this solution employs a time-dependent transportation network driven by hydrodynamics and a spatiotemporal racing mechanism. By constructing a road network model that dynamically decays with water depth and flow velocity, the time difference (evacuation margin) between the arrival time of flood danger and the shortest evacuation time is accurately calculated, and this physical time difference is nonlinearly mapped to the probability of successful evacuation. This approach upgrades risk assessment from traditional static superposition to dynamic survival probability correction, effectively solving the problem of distorted risk assessment caused by neglecting dynamic road network disruptions.

[0132] To address the lack of computational decision support in existing technologies, which makes it difficult to determine emergency repair priorities, this solution introduces a marginal contribution calculation method based on counterfactual inference. By constructing two sets of comparative scenarios—a baseline and an intervention scenario—the change in road network status after a specific road section is cleared is simulated, and the reduction in the number of stranded people resulting from this measure (i.e., the marginal contribution) is calculated. This approach assigns a calculated life-saving value to each damaged road section, solving the problem of lacking quantitative prioritization basis for emergency resource allocation and maximizing rescue benefits.

[0133] Example 11 describes a more robust and practically valuable scheduling algorithm that further addresses the practical application problems of potential volatility in single-moment evaluation and the lack of consideration for resource constraints in single-segment sorting.

[0134] Step 1101: Perform time window integration and robustness processing on the marginal contribution to eliminate the sensitivity of single-moment evaluation.

[0135] Marginal contribution ΔN is often calculated for a specific time t0. However, flood evolution is a dynamic process; a certain road segment may have a small contribution at time t0, but become a critical evacuation route at time t0+30 minutes. To improve the robustness of the assessment, this embodiment introduces a time window cumulative evaluation mechanism. Specifically, an evaluation time window [T] is set. start ,T end (For example, within 3 hours after the warning is issued). For each candidate road segment e, the system calculates its instantaneous marginal contribution ΔN(e,t) multiple times within a time window at fixed steps (e.g., 10 minutes), and then calculates the weighted cumulative value.

[0136] The formula is expressed as: ΔN robust (e)=∫ Tstart Tend w(t)·ΔN(e,t)dt;

[0137] Here, w(t) is the time weighting function. Typically, the closer to the arrival of the flood peak or the peak of personnel evacuation, the larger the weight w(t) is set. Through this time-domain integration, the evaluation error caused by data fluctuations at individual moments can be effectively smoothed out, and robust key road sections that truly have value for the entire rescue process can be screened out.

[0138] Step 1102: Construct a multi-segment combination optimization model with resource constraints to solve practical emergency dispatching problems.

[0139] In actual disaster relief efforts, engineering resources (such as excavators, pontoon bridge units, and sandbags) are often limited. Assuming the emergency command center has only K rescue teams (i.e., resource constraints), the objective is no longer simply ranking road sections, but rather selecting K sections from M damaged sections to reopen them, maximizing the reduction in the total number of trapped people.

[0140] Specifically, there are often non-linear coupling relationships between road segments. For example, road segment A and road segment B are connected in series; clearing either A or B alone will not open an evacuation route (the contribution is zero). Only by clearing both A and B simultaneously can significant rescue benefits be achieved. Simple single-segment sequencing methods cannot identify this combined effect.

[0141] Step 1103: Solve the multi-segment combination optimization model using heuristic algorithms or greedy strategies.

[0142] To solve the above combinatorial optimization problem (which is essentially a variant of the knapsack problem), this embodiment adopts one of the following two strategies:

[0143] Strategy 1 (Marginal Increment Greedy Method): Calculate the robust marginal contribution of each road segment individually, select the road segment e1 with the largest contribution and add it to the clearing set S; assuming e1 has been cleared, recalculate the conditional marginal contribution of the remaining road segments, select the road segment e2 with the largest increment and add it to S; repeat this process until K road segments are selected. This method can effectively capture the synergistic effect between road segments.

[0144] Strategy 2 (Genetic Algorithm / Simulated Annealing): Encode the rescue plan into a binary vector of length M (where K bits are 1), use the reduction in the total number of trapped people as the fitness function, and find the globally optimal combination plan through iterative search.

[0145] The system output includes not only a priority list, but also specific optimal road segment combinations, directly guiding the precise allocation of limited resources.

[0146] Example 12: According to one aspect of this application, the method further includes:

[0147] A multi-dimensional flood risk assessment method and system oriented towards people, transportation, and economy includes the following steps:

[0148] Step S1: Collect hydrological and meteorological data, topographic data, population distribution data, transportation network data, economic statistics data, and historical disaster data of the study area, and perform data cleaning, format conversion, and spatial registration;

[0149] Step S2: Based on a one- or two-dimensional hydrodynamic coupling model, simulate the inundation range, inundation depth, flow velocity, and inundation duration of floods with different return periods, and generate spatial distribution data during the flood evolution process;

[0150] Step S3: Based on population density distribution and flooding range, assess the exposure and vulnerability of personnel, and construct a personnel risk assessment model;

[0151] Step S4: Construct a traffic network topology model and identify key nodes and road segments, assess the impact of traffic facility flooding losses and traffic network connectivity, and establish a traffic risk assessment model.

[0152] Step S5: Identify economic exposure units, construct an economic loss assessment model, assess direct and indirect economic losses, and establish an economic risk assessment model.

[0153] Step S6: Determine the weights of different risk factors using the analytic hierarchy process (AHP), conduct a multidimensional risk assessment using the fuzzy comprehensive evaluation method, generate a comprehensive risk level zoning map, and visualize it.

[0154] The data collection in step S1 includes:

[0155] Step S11: Collect hydrological and meteorological data of the study area, including meteorological factors such as precipitation, evaporation, temperature, and wind speed;

[0156] Step S12: Collect terrain data, including digital elevation model (DEM) and surface feature data;

[0157] Step S13: Collect population distribution data, including information such as the number of permanent residents and age structure in different regions;

[0158] Step S14: Collect traffic network data, including information such as road network, bridge locations, and traffic flow;

[0159] Step S15: Collect economic statistics, including data related to economic activities such as regional GDP, industry distribution, and commercial and industrial assets;

[0160] Step S16: Collect historical disaster data, including the frequency, scale, and impact of past floods, as a reference for subsequent simulations and risk assessments;

[0161] Step S17: Perform data cleaning, format conversion, and spatial registration on the above data to ensure the spatiotemporal consistency of the data and the accuracy of subsequent analysis.

[0162] The flood inundation simulation in step S2 includes:

[0163] Step S21: Based on the digital elevation model (DEM) and hydro-meteorological data, the hydrological characteristics of the study area are modeled using hydrodynamic equations and a watershed model to simulate the flow characteristics under flood conditions with different return periods. The model uses a one-dimensional hydrodynamic model to describe the main channels within the watershed, and a two-dimensional hydrodynamic model to describe flood propagation in the riparian and urban areas.

[0164] Step S22: Simulate the inundation range, inundation depth, water flow velocity, and inundation duration of floods with different return periods: Based on historical meteorological data and future climate scenarios, combined with hydrological forecasts and precipitation predictions, simulate flood events with different return periods (such as 10 years, 50 years, 100 years, etc.) and calculate the inundation range, depth, water flow velocity, and duration of water flow.

[0165] Step S23: Through the flood process obtained by dynamic simulation, output the inundation area, inundation depth, water flow velocity and other data at each time step to form continuous spatial distribution data for subsequent risk assessment.

[0166] Step S24: Use actual observation data (such as water level, flow rate, etc.) to verify and validate the simulation results to ensure the accuracy of the model. Adjust and optimize the model based on the validation results to improve simulation accuracy.

[0167] The personnel risk assessment in step S3 includes:

[0168] Step S31: Based on population density distribution data, flood inundation depth, flow velocity, and human instability mechanism model of the study area, determine the number of people affected in each area. Calculate the number of exposed people in each area by spatially overlaying population density with exposed areas at different inundation depths and flow velocities.

[0169] Step S32: Assess the vulnerability of different groups by combining socioeconomic data, health status, age structure, and other vulnerability factors. Vulnerability factors include, but are not limited to, age (elderly, children), health status (people with chronic diseases or other illnesses), and socioeconomic status (low-income families or vulnerable groups).

[0170] Step S33: Construct a spatiotemporal dynamic population distribution model, introduce dynamic correction based on LBS (Location-Based Services) data, identify the spatial distribution pattern of the population in different time periods by analyzing location service data such as mobile phone signaling data and GPS trajectory data, and establish a dual-mode assessment system for the nighttime residential population and the daytime active population. This model can distinguish the exposure of people in different time periods at night and daytime, and dynamically adjust the number of statically exposed people calculated in step S31.

[0171] Step S34: Based on the dynamically corrected exposure, vulnerability, and flood hazard, construct a comprehensive personnel risk assessment model. This model uses the mathematical function R... 人员 =f(Dynamic Exposure, Vulnerability, Risk, Time Factor) assesses the risk level of individuals or groups under different flood disaster scenarios and time periods. The Dynamic Exposure combines static population distribution and LBS data correction results, and the Time Factor reflects the impact of the flood occurrence time on personnel risk. The risk factors are further weighted to obtain a comprehensive risk score that is dynamic in time and space.

[0172] Step S35: Establish a risk level classification standard for personnel. Based on the comprehensive risk score, divide the study area into five levels: extremely high risk, high risk, medium risk, low risk, and extremely low risk, and generate personnel risk zoning maps for different time periods (such as night, day, holidays, etc.).

[0173] The traffic risk assessment in step S4 includes:

[0174] Step S41: Construct the topology of the transportation network, abstracting transportation facilities such as roads, bridges, and tunnels into a network graph of nodes and edges; combine complex network theory to calculate network topology indicators, including node degree, betweenness centrality, and proximity centrality, and identify key nodes (such as important intersections, bridges, and tunnels) and key road segments (such as main roads and bottleneck sections) that are crucial to network connectivity; obtain attribute information such as traffic flow data, road grade, and traffic capacity to provide basic data for subsequent risk assessment;

[0175] Step S42: Establish inundation depth-loss rate curves for different types of transportation facilities, including ordinary roads, elevated roads, bridges, tunnels, and traffic signal facilities; based on the flood inundation simulation results, calculate the inundation depth of each transportation facility, and assess the degree of physical damage to the facilities using the loss rate curves; calculate the direct economic losses of the transportation facilities: DL 设施 =Σ(asset value i × loss rate i);

[0176] Step S43: Determine the capacity reduction of traffic facilities based on the flooding depth threshold, and establish a relationship model between flooding depth and capacity: when the flooding depth h < 0.3m, the capacity remains at 100%; when 0.3m ≤ h < 0.6m, the capacity drops to 50%; and when h ≥ 0.6m, the capacity is interrupted. Use network flow theory to analyze the impact of road network interruption on traffic flow distribution, calculate the changes in network connectivity indicators, assess the probability of critical path interruption and its impact on the overall accessibility of the traffic system, and identify the weak links in the traffic network.

[0177] Step S44: Based on historical disaster data and facility types, establish recovery time prediction models for different degrees of damage; consider the impact of factors such as emergency repair capabilities, resource allocation, and weather conditions on recovery time; calculate the transportation system function recovery curve and assess the time evolution of post-disaster transportation service levels.

[0178] Step S45: Establish a multi-dimensional traffic risk assessment model: R 交通 =α×R 设施损失 +β×R 网络中断 +γ×R 恢复时间 Where α, β, and γ are weighting coefficients; facility loss risk R 设施损失 Based on the calculation of direct economic losses, the network outage risk R 网络中断 Based on changes in connectivity metrics and flow loss calculations, the recovery time risk R... 恢复时间 Based on the duration of service interruption, and taking into account the vulnerability of the transportation system under different flood scenarios, a traffic risk level zoning map is generated.

[0179] The economic risk assessment in step S5 includes:

[0180] Step S51: Construct a multi-level economic exposure unit system, including a three-tiered exposure unit division at the macro level (industrial parks, commercial centers), the meso level (street blocks, communities), and the micro level (individual buildings, enterprises); establish a dynamic asset value assessment model based on the fusion of remote sensing imagery and POI data, combining multi-dimensional characteristics such as building area, building age, usage type, and location conditions to achieve accurate asset value estimation; introduce a time-dimensional asset value fluctuation model to consider the impact of economic cycles and seasonal changes on the value of different types of assets.

[0181] Step S52: Construct a multi-factor loss function considering water depth, flow velocity, inundation duration, and water pollution: DL=f(h,v,t,q)×AV; where h is water depth, v is flow velocity, t is inundation duration, q is water pollution index, and AV is asset value; establish damage mechanism models for different asset types, distinguishing between structural losses (building structure), content losses (equipment, inventory), and functional losses (business interruption), and establish corresponding loss rate functions for each; introduce building flood resilience assessment, and modify the loss rate function based on factors such as building materials, structural form, and protective measures;

[0182] Step S53: Construct a regional economic network model, establish the network topology of the economic system based on the supply chain relationship between enterprises and the industrial linkage matrix; use the cascading failure theory to analyze the propagation mechanism of flood impact in the economic network, and calculate the impact of node failure on the overall function of the network: IL=Σ(influence coefficient of failed node i × loss amplification factor of associated node j); establish a time-varying indirect loss propagation model, consider the time delay effect and attenuation characteristics of loss propagation, and dynamically evaluate the spatiotemporal evolution process of indirect economic losses.

[0183] Step S54: Establish a recovery cost model under the concept of better reconstruction, distinguishing between the cost of restorative reconstruction and the cost of resilience enhancement; construct a recovery time-cost trade-off model, considering the balance between rapid recovery and cost control; and introduce the impact of insurance coverage on actual recovery costs for correction.

[0184] Step S55: Establish a spatiotemporal dynamic economic risk assessment model: R 经济 (t)=ω1×R 直接 (t)+ω2×R 间接 (t)+ω3×R 恢复 (t)+ω4×R 机会成本 (t); where t is the time variable and ω is the dynamic weighting coefficient; introduce the economic resilience index to assess the economic system's ability to withstand shocks and recover: resilience index = (pre-disaster economic level - minimum economic level) / recovery time; construct a scenario-probability matrix, combine the probability of occurrence of different flood scenarios, calculate the expected economic loss and value at risk (VaR), and provide a probabilistic risk measure for risk management decisions;

[0185] Step S56: Generate a high-resolution economic risk heat map to achieve multi-scale risk visualization from macro-regions to micro-plots; establish an economic risk early warning threshold system and determine risk level classification standards in combination with regional economic affordability; provide differentiated risk management suggestions, including decision support information such as industrial layout optimization, insurance allocation strategies, and emergency fund reserves.

[0186] The multidimensional risk assessment in step S6 includes:

[0187] Step S61: Construct a multi-level risk indicator system, establishing a hierarchical structure model of target layer (comprehensive flood risk) - criterion layer (personnel risk, traffic risk, economic risk, ecological risk) - indicator layer (each sub-risk factor); use interval number hierarchical analysis to handle the uncertainty of expert judgment, obtain the possible value range of weights through interval number operations, and improve the scientific nature of weight determination; introduce time decay function and scenario sensitivity analysis to establish a dynamic weight adjustment mechanism: w i (t,s)=w i0 ×α(t)×β(s); where w i0 The basic weights are α(t), which is the time decay function, and β(s), which is the scenario adjustment factor.

[0188] Step S62: Transform the risk assessment results of each dimension into the basic probability allocation function (BPA) to construct the risk evidence body; use the improved Dempster-Shafer evidence theory to fuse multi-source risk information and handle cognitive uncertainty and random uncertainty in risk assessment; establish a conflict evidence handling mechanism, and when there is a conflict between the assessment results of different risk dimensions, use weighted average and reliability redistribution methods to make corrections.

[0189] Step S63: Construct a multi-agent risk assessment system, where each agent is responsible for the fuzzy assessment of a specific risk dimension, and reaches a consensus assessment result through a negotiation mechanism; extend the traditional fuzzy comprehensive assessment by adopting intuitionistic fuzzy set theory, and consider membership degree, non-membership degree and hesitation degree at the same time to more comprehensively characterize the fuzziness of risk; establish an adaptive membership function adjustment mechanism to dynamically optimize the membership function parameters based on historical disaster data and real-time monitoring information.

[0190] Step S64: Construct a hybrid model of convolutional neural network-long short-term memory network (CNN-LSTM) to learn the spatiotemporal feature patterns of multidimensional risk data; use an attention mechanism to identify key risk factors and sensitive areas to improve the accuracy and interpretability of risk assessment; establish a risk evolution prediction model to predict the risk change trend in future periods based on the current risk status;

[0191] Step S65: Establish a multi-scale risk aggregation framework to achieve cross-scale risk integration from grid units to administrative divisions; construct a risk spatial propagation model, considering the diffusion effect of risk in geographic space: R(x,y,t+Δt)=R(x,y,t)+Σ[D ij ×(R j -R i [)×Δt]; where D ij is the spatial diffusion coefficient; a risk threshold effect and a nonlinear superposition mechanism are introduced, triggering a risk amplification effect when multidimensional risks exceed the critical threshold;

[0192] Step S66: Use machine learning clustering algorithms (such as improved K-means, DBSCAN) to automatically classify risk levels to avoid human subjectivity; establish a dynamic update mechanism for risk levels, and automatically adjust the risk level boundaries by combining real-time monitoring data and early warning information; construct a confidence assessment system for risk levels and provide a confidence index for each risk level.

[0193] Step S67: Develop a three-dimensional risk landscape visualization system, using terrain elevation to represent risk intensity and achieve an intuitive display of risk spatial distribution; construct a spatiotemporal dynamic risk evolution animation to show the change process of risk in time and space dimensions; establish an interactive risk scenario analysis platform to support user-defined parameters for risk scenario simulation and comparative analysis; integrate virtual reality (VR) and augmented reality (AR) technologies to provide an immersive risk experience and emergency drill environment, enhancing risk awareness and emergency response capabilities.

[0194] To address the issue that existing technologies neglect the dynamic interaction between flood evolution and personnel evacuation, thus failing to accurately answer the question of whether escape is possible, this solution employs a time-dependent transportation network driven by hydrodynamics and a spatiotemporal racing mechanism. By constructing a road network model that dynamically decays with water depth and flow velocity, the time difference (evacuation margin) between the arrival time of flood danger and the shortest evacuation time is accurately calculated, and this physical time difference is nonlinearly mapped to the probability of successful evacuation. This approach upgrades risk assessment from traditional static superposition to dynamic survival probability correction, effectively solving the problem of distorted risk assessment caused by neglecting dynamic road network disruptions.

[0195] To address the lack of computational decision support in existing technologies, which makes it difficult to determine emergency repair priorities, this solution introduces a marginal contribution calculation method based on counterfactual inference. By constructing two sets of comparative scenarios—a baseline and an intervention scenario—the change in road network status after a specific road section is cleared is simulated, and the reduction in the number of stranded people resulting from this measure (i.e., the marginal contribution) is calculated. This approach assigns a calculated life-saving value to each damaged road section, solving the problem of lacking quantitative prioritization basis for emergency resource allocation and maximizing rescue benefits.

Claims

1. A multi-dimensional flood risk assessment method considering personnel, transportation, and economy, characterized in that, include: Acquire multi-source basic geographic data for the target area. Multi-source basic geographic data should include at least hydrological and meteorological data, topographic and geomorphological data, population distribution data, transportation network data, and economic asset data. Based on the hydrodynamic model, the evolution simulation of multi-source basic geographic data is carried out to generate flood spatiotemporal evolution data, which includes the time-varying inundation depth sequence and water flow velocity sequence. Based on flood spatiotemporal evolution data and population distribution data, the interaction characteristics between flood intensity and vulnerability of disaster-bearing bodies are analyzed, and the personnel safety risk index is calculated. Based on flood spatiotemporal evolution data and traffic network data, we analyze the attenuation characteristics of the physical state of traffic facilities and network capacity, and calculate the traffic operation risk index. Based on flood spatiotemporal evolution data and economic asset data, we analyze the exposure degree and value loss rate of different asset types and calculate the economic loss risk index. A multi-dimensional integrated evaluation based on personnel safety risk index, traffic operation risk index and economic loss risk index is conducted to obtain the multi-dimensional comprehensive flood risk assessment results for the target area. Among these, based on flood spatiotemporal evolution data and population distribution data, the interaction characteristics between flood intensity and the vulnerability of disaster-bearing bodies are analyzed, and a personnel safety risk index is calculated, including: For each assessment unit within the target area, acquire data on the time of arrival of personnel at risk; Based on the time-dependent transportation network, calculate the shortest evacuation time from the assessment unit to the nearest refuge site; Calculate the evacuation time margin, which represents the time difference between the arrival time of personnel in danger and the time required to complete the evacuation; Calculate the probability of successful evacuation based on the evacuation time margin; The static flood risk is dynamically corrected based on the probability of successful evacuation to obtain the personnel safety risk index; The formulas for calculating the evacuation time margin and the probability of successful evacuation are as follows: M(i,t)=T risk (i)-t-ΔT resp -T travel (i,t); P evac (i,t)=1 / 1+exp(-k·(M(i,t)-b)); Where M(i,t) represents the evacuation time margin of the i-th evaluation unit at time t; T risk (i) represents the time of arrival of personnel at risk in the assessment unit; ΔT resp Indicates the preset personnel response delay time; T travel (i,t) represents the shortest evacuation time calculated based on the time-dependent transportation network at time t; P evac (i,t) represents the probability of successful evacuation of the i-th evaluation unit at time t; k and b represent the preset probability mapping parameters; Before calculating the personnel safety risk index, the parameters of the probability mapping function are pre-calibrated: Collect data on personnel evacuation cases from historical flood disasters. The personnel evacuation case data includes historical evacuation time margins and corresponding evacuation outcome labels. Construct a logistic regression model with evacuation time margin as the independent variable and evacuation outcome label as the dependent variable; The logistic regression model was trained using maximum likelihood estimation based on personnel evacuation case data. The shape and position parameters of the probability mapping function were calculated and stored as preset parameters.

2. The method according to claim 1, characterized in that, Based on flood spatiotemporal evolution data and population distribution data, the interaction characteristics between flood intensity and vulnerability of disaster-bearing bodies are analyzed, and a personnel safety risk index is calculated, including: Extract the maximum inundation depth distribution from the spatiotemporal evolution data of floods; By spatially overlaying population distribution data with the maximum inundation depth distribution, the number of statically exposed people in the flooded area can be determined. Obtain the vulnerability factors of the population in the flooded area. The vulnerability factors include at least age structure characteristics and health status characteristics. The personnel safety risk index is calculated based on the number of statically exposed populations, the distribution of maximum inundation depths, and the population vulnerability factor.

3. The method according to claim 1, characterized in that, Based on flood spatiotemporal evolution data and transportation network data, this study analyzes the decay characteristics of the physical state of transportation facilities and network capacity, and calculates a traffic operation risk index, including: Construct a traffic network topology model based on traffic network data to identify key nodes and key road segments in the network; Obtain pre-constructed inundation depth-loss rate curves for transportation facilities; extract the maximum inundation depth at key nodes and key road sections from the spatiotemporal evolution data of the flood. The direct physical loss rate of critical nodes and critical road sections is calculated using the flooding depth-loss rate curve; Based on the direct physical loss rate and the traffic network topology model, the network connectivity decline index is calculated to obtain the traffic operation risk index.

4. The method according to claim 1, characterized in that, Based on flood spatiotemporal evolution data and economic asset data, this study analyzes the exposure levels and value loss rates of different asset types, and calculates an economic loss risk index, including: The economic asset data is divided into multi-level economic exposure units, and the asset value of each economic exposure unit is determined. Construct a multi-factor loss function. The input variables of the multi-factor loss function should include at least the flood depth, water flow velocity, and flood duration. Extract the inundation depth, flow velocity, and inundation duration characteristics of the spatiotemporal evolution data of the flood at each economically exposed unit; The projected economic loss for each economic exposure unit is calculated using a multi-factor loss function and asset value. The economic loss risk index is obtained by normalizing the expected economic loss amount.

5. The method according to claim 1, characterized in that, After generating the spatiotemporal evolution data of the flood, the following is also included: The target area is divided into grid cells according to a preset spatial resolution; For each grid cell within the target area, extract the moment when the flood depth sequence or water flow velocity sequence in the spatiotemporal evolution data of the flood first exceeds the preset personnel safety threshold to obtain the personnel danger arrival time data; For each road segment in the traffic network data, the moment when the flood depth sequence or water flow velocity sequence first exceeds the preset vehicle passage threshold is extracted from the flood spatiotemporal evolution data to obtain the road segment blockage arrival time data.

6. The method according to claim 5, characterized in that, Based on flood spatiotemporal evolution data and transportation network data, this study analyzes the decay characteristics of the physical state of transportation facilities and network capacity, and calculates a traffic operation risk index, including: Construct a time-dependent transportation network, where the edge attributes of the time-dependent transportation network include the passage capacity and travel time that change over time; For each road segment in a time-dependent transportation network, the dynamic traffic capacity of the road segment at different time steps is calculated based on the flood spatiotemporal evolution data of the flood depth sequence and water flow velocity sequence at that road segment using a preset capacity decay function. Based on the road blockage arrival time data, the dynamic traffic capacity value after the blockage arrival time is set to zero or a preset minimum value close to zero; Based on the time-dependent traffic network, the dynamic service level of the road network is calculated to obtain the traffic operation risk index.

7. The method according to claim 6, characterized in that, The preset capacity decay function is expressed as follows: C e (t)=C e0 ·exp(-a h ·(h e (t) / h0) p -a v ·(v e (t) / v0) q ); Among them, C e (t) represents the dynamic traffic capacity of the road segment at time t; C e0 The initial design capacity of the road segment is represented by exp(·); exp(·) represents an exponential function with the natural constant e as its base; h e (t) represents the submerged water depth of the road section at time t; v e (t) represents the water flow velocity of the road segment at time t; h0 represents the preset water depth characteristic reference value; v0 represents the preset flow velocity characteristic reference value; a h This indicates the preset weighting coefficient for the influence of water depth; a v This represents the preset weighting coefficient for the influence of flow velocity; p and q represent preset nonlinear exponential parameters.

8. The method according to claim 1, characterized in that, The static flood risk is dynamically adjusted based on the probability of successful evacuation to obtain the personnel safety risk index, which is calculated using the following formula: R(i,t)=H(i,t)·V(i)·(1-P evac (i,t)); Where R(i,t) represents the personnel safety risk index of the i-th assessment unit at time t; H(i,t) represents the flood hazard factor of the assessment unit at time t, the value of which is determined based on the inundation depth and flow velocity in the spatiotemporal evolution data of the flood; V(i) represents the population vulnerability factor of the assessment unit; P evac (i,t) represents the probability of successful evacuation of the evaluation unit at time t.

Images