A simulation data processing method for flood evacuation in a riverside mountainous city

Abstract

The application relates to the field of computer simulation and digital data processing, and discloses a simulation data processing method for flood evacuation in a riverside mountainous city. Multi-source data of a research area is collected and unified to the same spatial reference. A two-dimensional coupling hydrodynamic equation set is constructed. The Manning roughness is subjected to three terrain resistance corrections of slope, height difference and building, and is progressively superimposed in the order of slope resistance correction, height difference resistance correction and dense building resistance correction. Parameter optimization calibration is carried out based on historical flood mark water level. The water depth field and the flow velocity field are outputted, and the risk level is divided to obtain a dynamic risk layer sequence. A three-dimensional navigation road network is constructed, and the terrain step reduction coefficient and the water-related resistance reduction dynamic coefficient are mounted. The differential release and response delay modeling is carried out according to the heterogeneous population. The simulation solution is advanced by using a global synchronous clock. The path re-planning is triggered according to the real-time risk, and indexes are outputted. The simulation result deviating from the real risk avoidance problem caused by the isolation of each link in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of computer simulation and digital data processing technology, and in particular to a method for processing simulation data of flood evacuation in mountainous cities along rivers. Background Technology

[0002] Riverside mountain cities are built along the mountains and rivers, with low elevations in the riverside areas and high elevations in the inland terraces. Their road networks are dendritic and mainly consist of narrow alleys and stairways. During the flood season, floodwaters advance longitudinally along the main river channel and rapidly overflow into the built-up areas, creating a highly time-coupled dynamic with the evacuation process. In such scenarios, evacuation simulation data processing methods are a core component for emergency management departments in developing evacuation plans, assessing the capacity of key passageways, and optimizing the spatial layout of refuge facilities.

[0003] Existing evacuation simulation data processing solutions for this scenario typically use a one- or two-dimensional coupled hydrodynamic model of a plain city as the input to the disaster-causing field, and then superimpose a pedestrian evacuation model (such as a social force model, cellular automata, or basic agent model) based on a static reachability / inaccessibility binary judgment. However, through repeated backtracking and verification of historical flood trace data, UAV orthophotos, and measured water level sequences in typical riverside mountainous urban areas, it has been found that existing technical solutions have at least the following shortcomings: First, the one- or two-dimensional coupled hydrodynamic model distorts the energy loss of complex mountainous ground surfaces. The traditional unified static Manning roughness cannot characterize the three types of additional energy loss caused by drastic slope changes, vertical differences at terrace boundaries, and dense buildings, resulting in a systematic deviation between the simulated values ​​of water depth and flow velocity and the measured flood traces; Second, the evacuation simulation process is inadequate for complex terrain and... The joint response capability of dynamic water-related environments is insufficient. It ignores the continuous reduction of walking speed caused by steep slopes and steps, as well as the nonlinear reduction of walking speed caused by the product of water depth and flow velocity in low-to-medium risk water-related areas, which cannot be covered by a simple binary judgment of accessibility / inaccessibility. Thirdly, there is a lack of dynamic coupling between the temporal evolution of floods and the response of evacuation routes. It cannot reflect the continuous evolution of channels from unobstructed to blocked and then to blocked during the rise in water level, nor can it trigger path replanning at specific moments, resulting in abnormal personnel retention and paths deviating from the actual hazard avoidance patterns in the simulation results.

[0004] The three defects mentioned above are coupled with each other and amplify each other, which ultimately severely restricts the practical value of the simulation results and makes it difficult to provide reliable quantitative decision-making basis for emergency management departments. Summary of the Invention

[0005] The purpose of this invention is to provide a simulation data processing method for flood evacuation in mountainous cities along rivers, which solves the problem that the data processing links in the prior art are isolated from each other and lack coupling, resulting in the simulation results deviating from the actual disaster avoidance process.

[0006] To achieve the above objectives, this invention provides a method for processing simulation data on flood evacuation in riverside mountainous cities, comprising the following steps: Collect multi-source geographic and hydrological data of the study area, unify the collected data to the same spatial reference, and obtain standardized geographic data and hydrological boundary data; A one-dimensional Saint-Venant equation set and a two-dimensional shallow water equation set were constructed using standardized geographic data and hydrological boundary data. A one-dimensional and two-dimensional coupled hydrodynamic control equation set was established through lateral weir flow. The Manning roughness in the two-dimensional shallow water equation system is corrected for terrain resistance. The terrain resistance correction includes slope resistance correction, elevation difference resistance correction and dense building resistance correction. The corrections are progressively superimposed in the order of slope resistance correction, elevation difference resistance correction and dense building resistance correction to obtain the equivalent roughness. Using the measured water level at the historical flood mark verification point as a benchmark, the root mean square error is used as the objective function. The equivalent roughness is substituted into the one- and two-dimensional coupled hydrodynamic control equations. The parameters to be calibrated in the topographic resistance correction are optimized and calibrated in a partitioned manner to obtain the calibrated one- and two-dimensional coupled hydrodynamic control equations. The water depth field and velocity field output by the calibrated one- and two-dimensional coupled hydrodynamic control equations are divided into low-risk, medium-risk, and high-risk zones according to the hazard index to obtain a dynamic risk layer sequence. The hazard index is the product of the absolute values ​​of water depth and velocity. The risk level classification comprehensively considers the hazard index and the independent threshold conditions of water depth and velocity. A three-dimensional navigation road network is constructed based on the digital elevation model and building outlines in the standardized geographic data. The risk levels in the dynamic risk layer sequence are mapped to the navigation patch attributes of the three-dimensional navigation road network and a dynamic coefficient for wading resistance deceleration is attached. A terrain step speed reduction coefficient is attached to the navigation patch attributes. Based on the spatial distribution of buildings and roads in the standardized geographic data, differentiated spatial deployment is performed according to heterogeneous population types to obtain the initial location and response start time of each evacuation individual, and an early warning response delay is bound to each evacuation individual according to the response start time. The evacuation simulation is advanced by a global synchronous clock. In each step, the instantaneous step speed of each evacuation individual is calculated by reading the corresponding terrain step speed reduction coefficient and the water resistance deceleration dynamic coefficient based on the position of each evacuation individual on the navigation surface. When the water resistance deceleration dynamic coefficient steps to zero, global path replanning is triggered. The position of each evacuation individual is advanced according to the updated step speed and path until the simulation ends. The evacuation index is output after the simulation is completed.

[0007] The slope resistance correction is performed according to the following formula: in, The equivalent Manning roughness after terrain slope correction; The basic Manning roughness under a smooth surface condition; To calculate the terrain slope of the grid, which is the ratio of the elevation difference between grid center points to the horizontal distance; This is an empirical coefficient for slope correction, with a value ranging from 0.01 to 0.05 depending on the surface roughness.

[0008] The elevation difference resistance correction is performed according to the following formula: ; in, The composite equivalent roughness after superimposing the height difference effect; The equivalent Manning roughness after terrain slope correction; This represents the absolute terrain elevation difference between grid cells; The elevation difference morphological drag coefficient reflects the energy dissipation intensity of the impact of the waterfall; Calculate water depth locally for the grid; It is the acceleration due to gravity; The grid space step size is the direction of water flow.

[0009] The dense building resistance correction is performed according to the following formula: ; in, The effective Manning roughness of the mesh after building area correction; To calculate the building coverage rate within the grid, which is the ratio of the projected area of ​​buildings within the grid to the total area of ​​the grid; The composite equivalent roughness after superimposing the height difference effect; This refers to the extremely high water-resistant roughness corresponding to the solid area of ​​the building.

[0010] Specifically, using the measured water level at historical flood mark verification points as a benchmark, and employing the root mean square error as the objective function, the equivalent roughness is substituted into the one- or two-dimensional coupled hydrodynamic control equations to perform partitioned optimization calibration of the parameters to be calibrated in the topographic resistance correction. This includes: The parameters to be calibrated include the slope correction empirical coefficient, the elevation difference morphological resistance coefficient, and the equivalent water-resistant roughness of the building entity. Initial samples are generated in a three-dimensional parameter space using Latin hypercube sampling, driving a two-dimensional coupled hydrodynamic control equation system for forward calculation. After response surface fitting of the samples, a differential evolution algorithm is used for iterative optimization until the root mean square error converges. The root mean square error is calculated using the following formula: ; in, This represents the total number of measured floodmark verification points; This represents the peak water level calculated by the hydrodynamic model at the i-th verification point. Let be the measured flood mark water level corresponding to the i-th verification point.

[0011] The depth and velocity fields output by the calibrated one- or two-dimensional coupled hydrodynamic control equations are divided into low-risk, medium-risk, and high-risk zones according to hazard levels, specifically including: The hazard index D for low-risk areas is less than 0.4, and the water depth is... And the absolute value of the flow velocity The grid area; the hazard index D > 0.6 in the high-risk area, or the water depth or absolute value of flow velocity The grid area is the medium-risk area; the grid area is the remaining grid area excluding the low-risk area and the high-risk area. After classification, spatial filtering and geometric thinning are performed on the classification results to remove isolated pixels with an area smaller than a preset threshold.

[0012] Specifically, mapping the risk levels in the dynamic risk layer sequence to the navigation patch attributes of the three-dimensional navigation road network and attaching a dynamic coefficient for wading resistance deceleration includes: The value of the wading resistance deceleration dynamic coefficient is related to the risk level and time. In the high-risk area, the wading resistance deceleration dynamic coefficient jumps from 1.0 to 0.0 at the critical cutoff moment. In the medium-risk area, the wading resistance deceleration dynamic coefficient decays nonlinearly over time from the initial value when entering the medium-risk area to 0.25. In the low-risk area, the wading resistance deceleration dynamic coefficient decays gradually over time from the initial value of 1.0 when entering the low-risk area to 0.60.

[0013] Specifically, the terrain step speed reduction factor is attached to the navigation patch attributes, including: The terrain step speed reduction factor is obtained by multiplying the slope step speed reduction factor and the stairway step speed reduction factor; The slope speed reduction coefficient decreases monotonically as the spatial tilt angle of the navigation surface increases, and the stairway speed reduction coefficient decreases monotonically as the height difference between adjacent navigation surfaces increases.

[0014] In the evacuation simulation calculation driven by a global synchronous clock, the following sub-steps are executed sequentially at each step: Based on the location of each evacuated individual on the navigation patch, the risk level of the corresponding navigation patch at the current moment is extracted from the dynamic risk layer sequence, and the dynamic coefficient of water resistance deceleration is updated. Extract the terrain step speed reduction coefficient of each evacuation individual's location from the navigation patch attributes; The navigation patch with a dynamic coefficient of zero water resistance deceleration is temporarily removed from the three-dimensional navigation network, and a blocking signal is sent to the global path planner. Upon detecting a blocking signal, a global path replanning algorithm is triggered to replan an avoidance path for the blocked individual. The path cost function is obtained by weighting geometric distance, terrain speed reduction coefficient, and water resistance deceleration dynamic coefficient. For each evacuated individual, advance to the position at an instantaneous pace, apply an instantaneous throughput limit at the bottleneck channel based on the actual net width, and apply a social force repulsion model to individuals facing each other. Individuals who reach the safe exit are marked as having successfully evacuated and removed from the scene.

[0015] This includes implementing differentiated spatial targeting based on heterogeneous population types, specifically: The population in the study area was divided into two categories: the mobile population and the resident population. The mobile population is deployed using a uniform distribution algorithm, and the early warning response delay follows a normal distribution, with both the mean and standard deviation taking larger values. The permanent residents are deployed using a random distribution algorithm weighted by the residential function of the building. The early warning response delay follows a normal distribution, with the mean and standard deviation being the smallest values.

[0016] This invention discloses a data processing method for flood evacuation simulation in riverside mountainous cities. By integrating hydrodynamic simulation and personnel evacuation simulation into a unified data processing framework, the method first collects multi-source geographic and hydrological data of the study area and unifies them to the same spatial reference, constructs a one- or two-dimensional coupled hydrodynamic control equation set, then applies triple terrain resistance correction based on slope, elevation difference, and dense buildings to the Manning roughness coefficient, and performs parameter optimization and calibration based on historical floodwater levels to obtain a dynamic risk layer sequence. Simultaneously, a three-dimensional navigation road network is constructed, and terrain step speed reduction coefficient and wading resistance deceleration dynamic coefficient are attached to the navigation patch attributes. Combined with differentiated deployment of heterogeneous populations and early warning response delay, the simulation calculation is driven by a global synchronous clock, and path replanning is dynamically triggered according to real-time risks, ultimately outputting evacuation indicators. This invention, on the one hand, eliminates the distortion caused by the traditional uniform roughness factor, which cannot characterize the height difference of slopes and terraces and the additional energy consumption of dense buildings, through terrain resistance correction, thus significantly improving the accuracy of disaster sites. On the other hand, it corrects the neglect of the step speed attenuation of steep slopes, terraces and low-to-medium risk water crossing areas by the dual constraints of terrain step speed reduction and water resistance deceleration through the traditional binary passage judgment. Furthermore, through the real-time mapping of dynamic risk layers under global clock synchronization and the path replanning triggering mechanism, it realizes the dynamic coupling of channel state changes and evacuation path responses during the continuous evolution of floods, thereby effectively solving the problem of isolated data processing links and simulation results deviating from the actual disaster avoidance process in the existing technology. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0018] Figure 1 This is a flowchart of the simulation data processing method for flood evacuation in riverside mountainous cities according to the present invention.

[0019] Figure 2 This is a schematic diagram of the terrain resistance correction logic of the present invention.

[0020] Figure 3 This is a schematic diagram of the three-dimensional navigation road network construction and dual reduction mounting of the present invention.

[0021] Figure 4 This is a map showing the maximum inundation depth in riverside mountainous urban areas, based on the present invention.

[0022] Figure 5 This is a statistical comparison chart of the throughput of the key evacuation channels in this invention.

[0023] Figure 6 This is a flowchart of the steps in the simulation data processing method for flood evacuation in riverside mountainous cities according to the present invention. Detailed Implementation

[0024] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, but should not be construed as limiting the present invention.

[0025] Please refer to Figures 1 to 6 This invention provides a method for processing simulation data of flood evacuation in riverside mountainous cities, comprising the following steps: S101: Collect multi-source geographic and hydrological data of the study area, unify the collected data to the same spatial reference, and obtain standardized geographic data and hydrological boundary data; Specifically, the multi-source geographic and hydrological data includes: digital elevation model data, digital surface model data, building outline vector data, road centerline and street width vector data, underlying surface material zoning data, measured water level and flow time series from hydrological stations, historical flood mark location data, and post-disaster UAV orthophotos. Among them, the underlying surface material zoning data is divided into four categories: impermeable surfaces, paved surfaces, bare soil, and water surfaces.

[0026] After unifying the collected multi-source geographic and hydrological data to the national standard geodetic coordinate system and elevation datum, standardized geographic data and hydrological boundary data were obtained. The standardized geographic data includes digital elevation model data, digital surface model data, building outline vector data, road centerline and street width vector data, underlying surface material zoning data, and a topographic-building composite base map generated by overlaying building outline vectors and the digital elevation model. The hydrological boundary data includes time series of measured water levels and flows from hydrological stations, historical flood trace location data, and post-disaster UAV orthophotos, all based on the unified spatial datum. Spatial resampling with fault line preservation was performed on the digital elevation model and building outline vectors in the standardized geographic data to form a topographic-building composite base map of uniform resolution for the study area.

[0027] Preferably, the spatial resolution of the digital elevation model and the digital surface model is not less than 0.5m, and the scale of the building outline vector data is not less than 1:500, so as to ensure the detailed expression of features such as steep slopes, steps and narrow streets.

[0028] S102: Construct a one-dimensional Saint-Venant equation set and a two-dimensional shallow water equation set using standardized geographic data and hydrological boundary data, and establish a one-dimensional coupled hydrodynamic control equation set through lateral weir flow. Specifically, flood propagation in riverside mountainous cities involves two processes: longitudinal advancement of the main channel and planar diffusion within the built-up area. Flood movement in the main channel is essentially a one-dimensional unsteady, gradually varying flow, described by the Saint-Venant equations. These equations, consisting of continuity and momentum equations, are the standard mathematical tool in hydraulics for describing one-dimensional unsteady flow in open channels. Therefore, the one-dimensional Saint-Venant equations are used to obtain longitudinal water level and flow information. Flood overflow within the built-up area, however, exhibits a distinct two-dimensional diffusion characteristic, significantly influenced by factors such as topographic slope and building obstruction, requiring the use of two-dimensional shallow water equations for solution.

[0029] The one-dimensional Saint-Venant equations include the continuity equation and the momentum equation:

[0030]

[0031] Where: x: longitudinal distance along the river channel (m); t: calculation time step (s); A: cross-sectional area of ​​the water passage (m²) 2 Q: Flow rate of cross-sectional area (m³) 3 / s); h: free water surface elevation (m); q: lateral inflow per unit length (m) 2 / s), if none, then 0; g: gravitational acceleration, taken as 9.81 m / s². 2 ; : refers to the riverbed slope; : This refers to the decrease in friction ratio.

[0032] The two-dimensional shallow water equations are presented in total momentum form, including the continuity equation and the momentum equations in the x and y directions: Continuity equation:

[0033] Momentum equation:

[0034]

[0035] Where h is the local water depth (m); u and v are the velocity components in the x and y directions (m / s); and g is the gravitational acceleration (m / s²). 2 ); , The slope of the bottom slope; , This refers to the frictional slope. The frictional slope is characterized using the Manning formula, i.e. , where n is the Manning roughness.

[0036] The spatial coupling between the one-dimensional Saint-Venant equations and the two-dimensional shallow water equations employs a lateral weir connection. When the water level in the one-dimensional channel exceeds the bank elevation, a lateral overflow process is activated, and water diffuses from the one-dimensional channel to the two-dimensional block. The flow rate calculation at the coupling point follows the weir formula, ensuring the continuity of mass exchange between the one-dimensional and two-dimensional computational domains.

[0037] In the formula: Q is the overflow flow rate (m³ / s). 3 / s); m is the weir flow coefficient; B is the overflow section length (m); g is the gravitational acceleration (m / s²). 2 H represents the head of the weir crest (m), which is the difference between the river water level and the bank elevation.

[0038] Preferably, non-equidistant calculation cross-sections are arranged along the main channel in the one-dimensional river, and the cross-section density is increased at bends and cross-section contractions; the two-dimensional land area adopts an unstructured grid, which is adaptively increased at building boundaries, narrow streets and alleys, and stepped retaining walls, and the minimum grid side length should not be greater than 1.0m.

[0039] S103: Perform terrain resistance correction on the Manning roughness in the two-dimensional shallow water equation set. The terrain resistance correction includes slope resistance correction, elevation difference resistance correction and dense building resistance correction, and is progressively superimposed in the order of slope resistance correction, elevation difference resistance correction and dense building resistance correction to obtain the equivalent roughness. Specifically, in the two-dimensional shallow water equations established in step S102 above, the friction slope term includes the Manning roughness parameter n. Traditional schemes typically use a uniform static Manning roughness value across the entire computational domain. However, in riverside mountainous cities, dramatic slope variations, vertical elevation differences at terrace boundaries, and dense building layouts introduce three types of additional energy losses—amplified bottom shear friction, cascading impact resistance, and flow cross-sectional contraction resistance. A uniform static Manning roughness value cannot effectively characterize these additional losses. Therefore, it is necessary to perform terrain resistance correction on the Manning roughness value in the two-dimensional shallow water equations.

[0040] The two-dimensional computational mesh of the study area is divided into four categories based on the surface characteristics: continuous steep slope area, stepped retaining wall area, dense building area, and gently paved area. Corresponding correction sub-rules are then applied to each of the four categories. (a) Slope resistance correction For grids in areas of continuous steep slopes, the terrain gradient between grid center points is calculated, and the basic Manning roughness is corrected using a slope operator. In the formula: : Equivalent Manning roughness after terrain slope correction; : is the basic Manning roughness under smooth surface conditions; : To calculate the terrain slope of the grid (i.e., the ratio of the elevation difference between grid center points to the horizontal distance); : Slope correction empirical coefficient, with a value ranging from 0.01 to 0.05 depending on the surface roughness. When the grid is located at the edge of a terrace or a steep slope section, its equivalent roughness will be dynamically amplified, truly representing the sudden change and additional loss of kinetic energy in gravity-driven water flow in stepped and steep terrain.

[0041] (ii) Correction of elevation difference resistance For the stepped retaining wall area grid, the absolute terrain elevation difference between adjacent grids is extracted. Based on the principle of energy conservation, the physical elevation loss is converted into an equivalent resistance term: In the formula: : is the composite equivalent roughness after superimposing the height difference effect; : refers to the roughness coefficient after the aforementioned slope correction; : represents the absolute terrain elevation difference (m) between grids; : This is the elevation difference morphological drag coefficient, reflecting the energy dissipation intensity of the waterfall impact; : Calculates the water depth (m) locally within the grid. : is the acceleration due to gravity (taken as 9.81 m / s²). 2 ); : represents the grid spatial step size (m) in the direction of water flow. This equation establishes a mathematical relationship between the elevation difference ratio and the drag coefficient: when the flood level is lower than the height of the retaining wall, the formula can truly reflect the complete blocking effect of the elevation difference on the water flow; when the flood submerges the terrace, the formula reflects the reduction effect of the elevation difference on the flow capacity.

[0042] The above correction formula is derived based on the principle of converting local head loss into equivalent roughness: when water flows across a vertical height difference Δz formed by a terrace or retaining wall, the additional head loss caused by the impact of the cascading water and local backwater can be expressed as: Corresponding additional energy gradient: ; Will Incorporating Manning friction gradient:

[0043] The equivalent roughness square corresponding to this correction term can be obtained as follows: Based on the principle of energy gradient superposition, the friction gradient after slope correction is linearly superimposed with the additional energy gradient caused by the elevation difference. Right now: The above formula is obtained by taking the square root.

[0044] Dimensional verification: The dimensions of Δz, h, and Δx are all m, and the dimension of g is m. , Since it is dimensionless, the dimension of the correction term is (m^(4 / 3)·m) / (m·s^(−2)·m·m)=s²·m^(−2 / 3), which is consistent with... Since the dimensions are consistent, the two terms inside the square root are physical quantities with the same dimensions, and can be summed. The dimension of is s·m^(−1 / 3), consistent with the Manning roughness. The formula is dimensionally consistent and has a clear physical meaning.

[0045] (III) Dense Building Resistance Correction For grids in densely built-up areas, the base roughness is weighted and corrected according to the building coverage rate: In the formula: : The effective Manning roughness of the mesh after correction for building area; : This is used to calculate the building coverage within the grid, which is the ratio of the projected area of ​​buildings within the grid to the total area of ​​the grid; : This refers to the composite equivalent roughness after the aforementioned superimposed height difference effect; : The extremely high water-resistant roughness corresponding to the solid area of ​​the building.

[0046] In gently paved areas, the material's base roughness is used directly in the grid without further correction. In areas of continuous steep slopes, stepped retaining walls, and dense buildings, slope resistance correction, elevation difference resistance correction, and dense building resistance correction are applied respectively, and the correction sub-rules are superimposed in the following order: first, the material's base roughness is used... Based on the slope resistance correction, the equivalent Manning roughness after terrain slope correction is obtained. Secondly, Based on the baseline, a height difference resistance correction is performed to obtain the composite equivalent roughness after superimposing the height difference effect. Finally, with The dense building resistance correction is performed based on the baseline, that is, the basic roughness term in the dense building resistance correction formula is... Substituting the values, we obtain the equivalent roughness of the friction slope term in the final two-dimensional shallow water equation. When the same grid cell belongs to multiple topographic zones, the correction rules are applied sequentially in the order of slope correction, elevation difference correction, and dense building correction. When the grid cell belongs to only a single topographic zone, only the correction sub-rules corresponding to that zone are executed. Thus, the final equivalent roughness is... Determined by slope, elevation difference, building coverage, and materials. S104: Using the measured water level of the historical flood mark verification point as a benchmark, the root mean square error is used as the objective function. The equivalent roughness is substituted into the one- or two-dimensional coupled hydrodynamic control equation set. The parameters to be calibrated in the topographic resistance correction are calibrated by partitioning and optimization to obtain the calibrated one- or two-dimensional coupled hydrodynamic control equation set. Specifically, after completing the terrain resistance correction in step S103 above, the obtained equivalent roughness... It contains several parameters that need to be calibrated using measured data. These parameters include empirical coefficients for slope correction. Elevation Difference Shape Resistance Coefficient Equivalent water-resistant roughness of building structure Set the initial physical constraint range for the above-mentioned parameters to be calibrated. ∈[0.01, 0.05], ∈[10.0, 50.0], ∈[0.10, 0.30].

[0047] Using the measured water levels at historical flood verification points as a benchmark, and the root mean square error as the core objective function for evaluating simulation accuracy:

[0048] In the formula: : This represents the total number of measured floodmark verification points; : The peak water level (m) calculated by the hydrodynamic model at the i-th verification point; : is the measured flood mark water level (m) corresponding to the i-th verification point.

[0049] Initial samples were generated in the three-dimensional parameter space using Latin hypercube sampling, which then drove the one- and two-dimensional coupled hydrodynamic control equations for forward calculation. The samples were fitted to the response surface, and the differential evolution algorithm was used to iteratively optimize them. When the objective function approached the global minimum and the global flooding boundary highly matched the post-disaster UAV orthophoto, the set of parameters was locked as the optimal calibration result for the study area.

[0050] Preferably, the model accuracy evaluation adopts three indicators: absolute error of water level AE, Nash efficiency coefficient NSE, and coefficient of determination R². When the three indicators NSE>0.75, coefficient of determination R²>0.85, and absolute error of water level AE≤1.0m are satisfied simultaneously, the model is judged to reach the excellent level and can proceed to the next step.

[0051] S105: The depth field and velocity field output by the calibrated one- or two-dimensional coupled hydrodynamic control equations are divided into low-risk, medium-risk, and high-risk zones according to the hazard index to obtain a dynamic risk layer sequence. The hazard index is the product of the absolute values ​​of the depth and the velocity. The risk level classification takes into account the hazard index and the independent threshold conditions of the depth and velocity. Specifically, the two-dimensional coupled hydrodynamic control equations calibrated in step S104 output the two-dimensional depth field H(x,y,t) and velocity field V(x,y,t) for each time step t. A comprehensive hazard index is calculated for each grid. V represents the flow velocity (m / s), and D represents the hazard index (m² / s). Furthermore, to avoid confusion between hazard levels caused by the same product index corresponding to two different hydraulic states—"slow and deep" and "shallow and fast"—the risk level classification considers not only the hazard index D but also the independent threshold conditions of water depth H and the absolute value of flow velocity |V|.

[0052] Based on the critical conditions for human instability while wading, D(x,y,t) is divided into three levels: Low-risk area: D < 0.4 and Characterizes shallow, slow-flowing water conditions; High-risk area: or H≥1.5 m, or |V|≥2.5 m / s, characterizing the dangerous state of deep water jet, narrow jet, or any single property of water depth or velocity exceeding the limit; Medium-risk zone: The remaining grids, excluding the low-risk and high-risk zones, represent the diffuse flow state with significant drag force.

[0053] Among them, the water depth thresholds of 1.0 m and 1.5 m correspond to the submersion depths above the waist and chest of adults, respectively, and the flow velocity thresholds of 1.5 m / s and 2.5 m / s correspond to the critical flow velocity ranges for human instability in wading stability studies. Based on the above two-dimensional discrimination, the hazard index D is the same for both the "slow and deep" (H=1.0 m, |V|=0.4 m / s, D=0.4 m² / s) and "shallow and fast" (H=0.1 m, |V|=4.0 m / s, D=0.4 m² / s) conditions. However, the latter is classified as a high-risk area because its flow velocity exceeds 2.5 m / s, while the former is classified as a medium-risk area. This reasonable distinction in hazard levels overcomes the problem of conflation between the two conditions under a single product index.

[0054] Spatial filtering and geometric thinning are performed on the classification results to remove isolated pixels and pseudo-risk patches with an area smaller than the preset threshold (4 grids by default), resulting in a dynamic risk layer sequence with closed boundaries and regular patches.

[0055] S106: Construct a three-dimensional navigation road network based on the digital elevation model and building outlines in the standardized geographic data, map the risk level in the dynamic risk layer sequence to the navigation patch attributes of the three-dimensional navigation road network and attach a dynamic coefficient for wading resistance deceleration, and attach a terrain step speed reduction coefficient to the navigation patch attributes; Specifically, using the digital elevation model and building outline vectors from standardized geographic data as input, an initial two-dimensional passable surface is generated by subtracting the building footprint. Delaunay triangulation is performed on this two-dimensional surface, and elevation attributes are extracted from the digital elevation model according to the grid centroid, thus elevating the two-dimensional surface into a three-dimensional navigation surface. Dynamic physical stitching is performed at the junctions of the riverbank floodplain, steps, and ramps, i.e., when the projected distance between two adjacent surfaces is less than the set tolerance (default 0.3m) and the elevation difference is less than the set value (default 0.5m), a connected edge is forcibly established. Fragmented surfaces with an area less than the set value (default 0.5m²) are merged or filtered out. Finally, a physically connected three-dimensional navigation road network is generated throughout the study area.

[0056] On the three-dimensional navigation network, three special elements are further identified and labeled: narrow bottleneck channels, which limit the maximum flow rate per unit time according to the actual net width; safe escape exits, which are inland highland nodes located above the extreme flood level, serving as the absolute endpoint of the evacuation simulation; and inaccessible obstacles, namely building walls and deep water bodies, which are set as rigid boundaries with zero normal velocity.

[0057] (a) Loading of dynamic coefficient of water resistance deceleration From the dynamic risk layer sequence obtained in step S105, extract the risk level corresponding to each navigation patch at different times. Assign each navigation patch a velocity reduction coefficient that evolves over time, with a value range of [0.0, 1.0], where 1.0 represents normal walking speed efficiency and 0.0 represents complete physical connectivity disruption. Label each patch according to its risk level: The velocity reduction factor in the high-risk area jumps from 1.0 to 0.0 at the critical cutoff moment, characterizing the instantaneous cutoff effect of the deep-water rapid on the key escape route; The velocity reduction factor in the medium-risk zone decreases nonlinearly over time from its initial value when entering the medium-risk zone to 0.25, indicating that the fluid drag force is approaching the human body's overturning resistance limit. The velocity reduction factor in the low-risk zone gradually decreases from an initial value of 1.0 when entering the low-risk zone to 0.60 over time, indicating that only wading viscous resistance needs to be overcome.

[0058] (ii) Mounting of vehicles with terrain-based pace reduction factor For each navigation patch in the 3D navigation network, its spatial tilt angle and the height difference between adjacent patches are extracted in real time, and the terrain step speed reduction factor is calculated using the following formula: in, This is the slope-rate reduction function, with a value range of [0.4, 1.0], which decreases as the slope increases; This is the terrain step speed reduction function, which decreases as the elevation difference between adjacent patches increases. The terrain step speed reduction coefficient and the wading resistance deceleration dynamic coefficient are both loaded into the navigation patch attribute library for use in subsequent simulation calculation steps.

[0059] Preferably, the slope-step-rate reduction function can be a piecewise linear function: 1.0 is used when the angle is ≤5°, and 1.0 is used when the angle is <5°. For angles ≤15°, take 1.0−0.02·( -5), 15°< For angles ≤25°, take 0.8−0.03·( -15), When the angle is >25°, take 0.5; the step speed reduction function of the stairway can be taken in stages: take 1.0 when the height difference between adjacent surfaces is ≤0.15m, take 0.9 when the height difference between adjacent surfaces is ≤0.30m and take 0.75 when the height difference between adjacent surfaces is >0.30m.

[0060] S107: Based on the spatial distribution of buildings and roads in the standardized geographic data, perform differentiated spatial deployment according to heterogeneous population types, obtain the initial location and response start time of each evacuation individual, and bind an early warning response delay to each evacuation individual according to the response start time. Specifically, the population in the study area is divided into two categories: mobile population and resident population, and different spatial distribution algorithms and response delay models are assigned to each category: The mobile population consists of people with low environmental awareness and weak crisis awareness. The initial spatial deployment uses a uniform distribution algorithm to sprinkle points on the commercial activity area. The early warning response delay follows a normal distribution, with both the mean and standard deviation taking large values, indicating that this population has greater uncertainty in disaster response time. The resident population consists of people familiar with the terrain and with strong risk prediction capabilities. The initial spatial deployment adopts a random distribution algorithm weighted by the residential function of the building. The early warning response delay follows a normal distribution, with both the mean and standard deviation taking smaller values, indicating that this type of population is more familiar with the environment and responds more quickly.

[0061] Each evacuated individual is independently sampled to obtain its response initiation time and initial position; during the initiation phase, the individual remains in place and does not participate in pathfinding calculations.

[0062] S108: The evacuation simulation calculation is advanced by a global synchronous clock. In each step, the instantaneous step speed of the individual is calculated by reading the corresponding terrain step speed reduction coefficient and the water resistance deceleration dynamic coefficient based on the position of each evacuation individual on the navigation surface. When the water resistance deceleration dynamic coefficient steps to zero, global path replanning is triggered. The position of each evacuation individual is advanced according to the updated step speed and path until the simulation ends. Specifically, a global synchronization clock is introduced to drive the simulation engine forward in extremely small steps (0.5s by default); within each step, the following sub-steps are executed sequentially: (1) Status scanning and coefficient refresh: Based on the location of each evacuated individual in the navigation patch, extract the risk level of the corresponding navigation patch at the current moment from the dynamic risk layer sequence, and refresh the dynamic coefficient of water resistance deceleration; extract the terrain step speed reduction coefficient of each evacuated individual in the navigation patch attribute.

[0063] (2) Connectivity determination and blocking signal transmission: The navigation patch with zero dynamic coefficient of water resistance deceleration is temporarily removed from the three-dimensional navigation road network and a blocking signal is sent to the global path planner.

[0064] (3) Global path replanning: Once a blocking signal is detected, a global path replanning algorithm (such as the A* algorithm or the Dijkstra algorithm) is triggered to find a new optimal risk avoidance path for the blocked individual. The path cost function is obtained by weighting the geometric distance, the terrain speed reduction coefficient and the water resistance deceleration dynamic coefficient.

[0065] (4) Individual dynamics calculation: For each evacuated individual, its instantaneous output gait velocity is calculated using the following formula:

[0066] In the formula: : represents the actual microscopic walking speed of individual i at time t during evacuation. : represents the actual microscopic walking speed of the evacuated individual at time i. : This is the dynamic velocity reduction factor, and its value range is between 0.0 and 1.0. : The spatial grid patch where evacuated individual i is located at time t, used to index the hazard level.

[0067] (5) Local avoidance and congestion relief: At the bottleneck channel, an instantaneous throughput limit is applied according to the actual net width, and individuals exceeding the limit form a queue in front of the channel; a social force repulsion model is applied to individuals facing each other to avoid clipping through.

[0068] (6) Individual status update: Advance the individual's position by the product of instantaneous step speed and time step. If the individual reaches the safe exit, it is marked as successfully evacuated and removed from the scene.

[0069] The simulation terminates when all individuals reach the safe exit or the maximum simulation duration is reached.

[0070] S109: Output evacuation indicators after the simulation is completed.

[0071] Specifically, the following key metrics will be output after the simulation is completed: A. Spatial water depth field, current velocity field, and hazard field at each moment; B. Overall and zoned evacuation success rate, average evacuation time, and average route extension rate; C. The proportion of evacuation load borne by each safety exit; D. Time-series traffic flow and peak congestion levels at key bottleneck channels; E. Number of route replanning attempts triggered by high-risk area blockades and the number of people affected.

[0072] Example 1: Complete implementation process for a typical riverside mountainous city area: This embodiment describes the complete implementation process of the present invention in a typical riverside mountainous urban area. The study area is located on the banks of a river, possessing both historical commercial and residential attributes, and has a typical hillside-terrace-stepped composite spatial structure. The total area covered by the study is approximately 0.995 km².

[0073] Data preparation: Data was collected from the study area using a 12.5m resolution digital elevation model (obtained via ALOS PALSAR (Advanced Land Observation Satellite Phased Array L-band Synthetic Aperture Radar) satellite), 1:500 scale building outline vector data (approximately 1200 buildings), street centerline and net width vector data, and underlying surface material zoning raster (four categories: impermeable surface, paved surface, bare soil, and water surface). Nearly 20 years of measured water level and flow time series from nearby hydrological stations in the lower reaches of the Jialing River (including Beibei Station, Cuntan Station, and Ciqikou Station) were also collected, along with 30 historical floodmark verification points and three phases of post-disaster UAV orthophotos. All data were unified to the CGCS2000 coordinate system and the 1985 National Height Datum, and resampled to a 2m computational grid resolution, forming a topographic-building composite base map of the study area with approximately 3.5 million grids.

[0074] Equation system construction: 145 non-equidistant cross-sections are arranged along the Jialing River in one-dimensional channel, with emphasis on densification at bends and cross-sectional contractions; the two-dimensional land area uses an unstructured grid, adaptively densified at building boundaries, narrow streets, and stepped retaining walls, with a minimum grid side length of 1.0m and a maximum of 4.0m. 82 lateral weir connection sections are arranged along the bank slope at the one-dimensional and two-dimensional coupling points.

[0075] Terrain resistance correction and calibration: The 2m grid was automatically divided into four categories based on slope, elevation difference, building coverage, and material. The empirical coefficient for slope correction, the elevation difference morphological resistance coefficient, and the equivalent water-blocking roughness of the building entity were optimized in the three-dimensional parameter space using Latin hypercube sampling and differential evolution algorithm. This process drove 42 rounds of forward calculation of the coupled one- and two-dimensional hydrodynamic control equations, ultimately locking in the following parameter combinations: empirical coefficient for slope correction = 0.028, elevation difference morphological resistance coefficient = 27.5, and equivalent water-blocking roughness of the building entity = 0.18. The root mean square error of 30 verification points decreased from 0.63m under the traditional uniform roughness to 0.16m, improving simulation accuracy by 74.6%. The Nash efficiency coefficient (NSE) was 0.83, and the coefficient of determination (R²) was 0.90, achieving an excellent level.

[0076] Output for the disaster site: An extreme scenario input was a once-in-a-century flood (peak water level of approximately 194.5m in the study area section), with a simulation duration of 4800s and an output step size of 10s. The hazard index (the product of water depth and absolute flow velocity) and independent thresholds for water depth and flow velocity were used to divide the entire grid into three levels: low, medium, and high. The classification results were then filtered and thinned using an area threshold of 4 grids to obtain a dynamic risk layer of 480 frames.

[0077] 3D navigation mesh construction: After deducting the building footprint, Delaunay triangulation was performed on the 0.42 km² passable surface of the study area, resulting in approximately 126,000 3D navigation patches; the stitching tolerance was 0.3 m, the elevation difference tolerance was 0.5 m, the debris filtering threshold was 0.5 m², and 8 safety exits (distributed in inland highlands above 210 m elevation) and 26 narrow bottleneck passages (the narrowest clear width was 1.2 m) were identified.

[0078] Dual reduction and scaling: In high-risk areas, critical cutoff times are triggered as if under high risk for 60 seconds; in medium-risk areas, the dynamic coefficient of water resistance deceleration decreases non-linearly to 0.25; in low-risk areas, the dynamic coefficient of water resistance deceleration decreases gradually to 0.60; terrain step speed reduction uses a piecewise function. The global terrain step speed reduction coefficient and the dynamic coefficient of water resistance deceleration are scaled to 126,000 navigation patches.

[0079] Population deployment: Based on the average daily pedestrian flow distribution in the study area, 8,500 people were deployed to the mobile group (mean response delay 300s, standard deviation 90s) and 3,200 people were deployed to the resident group (mean response delay 120s, standard deviation 40s), for a total of 11,700 people; individual baseline walking speed was sampled according to gender-age distribution, with values ​​ranging from 0.9 to 1.55 m / s.

[0080] Simulation results: The simulation step size was 0.5s, and the global synchronization clock advanced by 4800s; a total of 6 global path replanning events were triggered (corresponding to the blocking times of multiple key roads in the core area). Key performance indicators (KPIs) after the simulation were output: overall evacuation success rate was 94.4%, average evacuation time was 1832s, and average path extension rate was 1.42; among the 8 safety exits, one exit located at an elevation of 218m bore 41.5% of the evacuation load; some nodes experienced congestion peaks between 780 and 1140s, with a peak density of 4.8 people / m².

[0081] The above indicators can be directly used for the quantitative evaluation of emergency plans in the study area. These include identifying high loads at specific exits, peak density exceeding safety thresholds at some bottleneck channels, and the marginal impact of path replanning in some directions on the evacuation time of mobile groups. All of these can point to specific recommendations for channel modification and plan optimization.

[0082] Example 2: Implementation and promotion in a larger-scale riverside area: This embodiment describes the extension of the invention to a riverside mountainous urban area with similar spatial characteristics but on a larger scale. The example focuses on a riverside area (approximately 3.1 km²) near the confluence of River A and River B, which has a more complex road network structure and a larger population load.

[0083] Differentiated configuration: 220 cross-sections are arranged along the main channel of the Yangtze River in one dimension, and 96 cross-sections are arranged along the tributaries; approximately 7.8 million two-dimensional land grids are used; the parameter calibration values ​​are: slope correction empirical coefficient 0.031, elevation difference morphological resistance coefficient 24.8, and equivalent water-blocking roughness coefficient of building entity 0.22; there are 15 safety exits and 58 bottleneck channels; a total population of approximately 26,800 people are deployed.

[0084] Simulation results: Under the same once-in-a-century extreme flood scenario, the overall evacuation success rate was 91.7%, the average evacuation time was 2310s, the average path extension rate was 1.56, and the 5 high-level exits bore 63.2% of the evacuation load.

[0085] Performance Comparison: Compared to the traditional scheme using a unified roughness factor plus static reachability / unreachability binary judgment, the present invention identifies three critical failure paths that the traditional scheme could not identify, and provides the specific trigger time for path replanning. This embodiment demonstrates that the present invention has good applicability and consistency in cross-regional and cross-scale scenarios.

[0086] This invention integrates hydrodynamic simulation and personnel evacuation simulation into a unified data processing framework, sequentially executing steps S101 to S109. It corrects the Manning roughness using triple terrain resistance factors (slope, elevation difference, and dense buildings) and performs parameter optimization calibration based on historical flood levels. It attaches terrain step speed reduction coefficients and wading resistance deceleration dynamic coefficients to the three-dimensional navigation patch attributes. The simulation is driven by a global synchronization clock, and path replanning is dynamically triggered based on real-time risks. This achieves dynamic coupling between disaster site simulation and evacuation simulation, effectively solving the problem in existing technologies where isolated data processing stages lead to simulation results deviating from the actual disaster avoidance process.

[0087] The above-disclosed embodiments are merely one or more preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art can understand that all or part of the processes for implementing the above embodiments and equivalent changes made in accordance with the claims of this application still fall within the scope of this application.

Claims

1. A method for processing simulation data on flood evacuation in riverside mountainous cities, characterized in that, Includes the following steps: Collect multi-source geographic and hydrological data of the study area, unify the collected data to the same spatial reference, and obtain standardized geographic data and hydrological boundary data; A one-dimensional Saint-Venant equation set and a two-dimensional shallow water equation set were constructed using standardized geographic data and hydrological boundary data. A one-dimensional and two-dimensional coupled hydrodynamic control equation set was established through lateral weir flow. The Manning roughness in the two-dimensional shallow water equation system is corrected for terrain resistance. The terrain resistance correction includes slope resistance correction, elevation difference resistance correction and dense building resistance correction. The corrections are progressively superimposed in the order of slope resistance correction, elevation difference resistance correction and dense building resistance correction to obtain the equivalent roughness. Using the measured water level at the historical flood mark verification point as a benchmark, the root mean square error is used as the objective function. The equivalent roughness is substituted into the one- and two-dimensional coupled hydrodynamic control equations. The parameters to be calibrated in the topographic resistance correction are optimized and calibrated in a partitioned manner to obtain the calibrated one- and two-dimensional coupled hydrodynamic control equations. The water depth field and velocity field output by the calibrated one- and two-dimensional coupled hydrodynamic control equations are divided into low-risk, medium-risk, and high-risk zones according to the hazard index to obtain a dynamic risk layer sequence. The hazard index is the product of the absolute values ​​of water depth and velocity. The risk level classification comprehensively considers the hazard index and the independent threshold conditions of water depth and velocity. A three-dimensional navigation road network is constructed based on the digital elevation model and building outlines in the standardized geographic data. The risk levels in the dynamic risk layer sequence are mapped to the navigation patch attributes of the three-dimensional navigation road network and a dynamic coefficient for wading resistance deceleration is attached. A terrain step speed reduction coefficient is attached to the navigation patch attributes. Based on the spatial distribution of buildings and roads in the standardized geographic data, differentiated spatial deployment is performed according to heterogeneous population types to obtain the initial location and response start time of each evacuation individual, and an early warning response delay is bound to each evacuation individual according to the response start time. The evacuation simulation is advanced by a global synchronous clock. In each step, the instantaneous step speed of each evacuation individual is calculated by reading the corresponding terrain step speed reduction coefficient and the water resistance deceleration dynamic coefficient based on the position of each evacuation individual on the navigation surface. When the water resistance deceleration dynamic coefficient steps to zero, global path replanning is triggered. The position of each evacuation individual is advanced according to the updated step speed and path until the simulation ends. The evacuation index is output after the simulation is completed.

2. The method for processing simulation data on flood evacuation in riverside mountainous cities as described in claim 1, characterized in that, The slope resistance correction is performed according to the following formula: ； in, The equivalent Manning roughness after terrain slope correction; The basic Manning roughness under a smooth surface condition; To calculate the terrain slope of the grid, which is the ratio of the elevation difference between grid center points to the horizontal distance; This is an empirical coefficient for slope correction, with a value ranging from 0.01 to 0.05 depending on the surface roughness.

3. The method for processing simulation data on flood evacuation in riverside mountainous cities as described in claim 2, characterized in that, The elevation difference resistance correction is performed according to the following formula: ； in, The composite equivalent roughness after superimposing the height difference effect; The equivalent Manning roughness after terrain slope correction; This represents the absolute terrain elevation difference between grid cells; The elevation difference morphological drag coefficient reflects the energy dissipation intensity of the impact of the waterfall; Calculate water depth locally for the grid; It is the acceleration due to gravity; The grid space step size is the direction of water flow.

4. The method for processing simulation data on flood evacuation in riverside mountainous cities as described in claim 3, characterized in that, The dense building resistance correction is performed according to the following formula: ； in, The effective Manning roughness of the mesh after building area correction; To calculate the building coverage rate within the grid, which is the ratio of the projected area of ​​buildings within the grid to the total area of ​​the grid; The composite equivalent roughness after superimposing the height difference effect; This refers to the extremely high water-resistant roughness corresponding to the solid area of ​​the building.

5. The method for processing simulation data on flood evacuation in riverside mountainous cities as described in claim 1, characterized in that, Using the measured water level at historical floodmark verification points as a benchmark, and employing the root mean square error as the objective function, the equivalent roughness is substituted into the one- or two-dimensional coupled hydrodynamic control equations to perform partitioned optimization calibration of the parameters to be calibrated in the topographic resistance correction. Specifically, this includes: The parameters to be calibrated include the slope correction empirical coefficient, the elevation difference morphological resistance coefficient, and the equivalent water-resistant roughness of the building entity. Initial samples are generated in a three-dimensional parameter space using Latin hypercube sampling, driving a two-dimensional coupled hydrodynamic control equation system for forward calculation. After response surface fitting of the samples, a differential evolution algorithm is used for iterative optimization until the root mean square error converges. The root mean square error is calculated using the following formula: ； in, This represents the total number of measured floodmark verification points; This represents the peak water level calculated by the hydrodynamic model at the i-th verification point. Let be the measured flood mark water level corresponding to the i-th verification point.

6. The method for processing simulation data on flood evacuation in riverside mountainous cities as described in claim 1, characterized in that, The depth and velocity fields output by the calibrated one- or two-dimensional coupled hydrodynamic control equations are divided into low-risk, medium-risk, and high-risk zones according to hazard levels, specifically including: The hazard index D for low-risk areas is less than 0.4, and the water depth is... And the absolute value of the flow velocity The grid area; the hazard index D > 0.6 in the high-risk area, or the water depth or absolute value of flow velocity The grid area is the medium-risk area; the grid area is the remaining grid area excluding the low-risk area and the high-risk area. After classification, spatial filtering and geometric thinning are performed on the classification results to remove isolated pixels with an area smaller than a preset threshold.

7. The method for processing simulation data on flood evacuation in riverside mountainous cities as described in claim 1, characterized in that, Mapping the risk levels in the dynamic risk layer sequence to the navigation patch attributes of the three-dimensional navigation road network and attaching a dynamic coefficient for wading resistance deceleration specifically includes: The value of the wading resistance deceleration dynamic coefficient is related to the risk level and time. In the high-risk area, the wading resistance deceleration dynamic coefficient jumps from 1.0 to 0.0 at the critical cutoff moment. In the medium-risk area, the wading resistance deceleration dynamic coefficient decays nonlinearly over time from the initial value when entering the medium-risk area to 0.

25. In the low-risk area, the wading resistance deceleration dynamic coefficient decays gradually over time from the initial value of 1.0 when entering the low-risk area to 0.

60.

8. The method for processing simulation data of flood evacuation in riverside mountainous cities as described in claim 1, characterized in that, The terrain step speed reduction factor is attached to the navigation patch attributes, specifically including: The terrain step speed reduction factor is obtained by multiplying the slope step speed reduction factor and the stairway step speed reduction factor; The slope speed reduction coefficient decreases monotonically as the spatial tilt angle of the navigation surface increases, and the stairway speed reduction coefficient decreases monotonically as the height difference between adjacent navigation surfaces increases.

9. The method for processing simulation data on flood evacuation in riverside mountainous cities as described in claim 1, characterized in that, In the evacuation simulation solution driven by a globally synchronized clock, the following sub-steps are executed sequentially at each step: Based on the location of each evacuated individual on the navigation patch, the risk level of the corresponding navigation patch at the current moment is extracted from the dynamic risk layer sequence, and the dynamic coefficient of water resistance deceleration is updated. Extract the terrain step speed reduction coefficient of each evacuation individual's location from the navigation patch attributes; The navigation patch with a dynamic coefficient of zero water resistance deceleration is temporarily removed from the three-dimensional navigation network, and a blocking signal is sent to the global path planner. Upon detecting a blocking signal, a global path replanning algorithm is triggered to replan an avoidance path for the blocked individual. The path cost function is obtained by weighting geometric distance, terrain speed reduction coefficient, and water resistance deceleration dynamic coefficient. For each evacuated individual, advance to the position at an instantaneous pace, apply an instantaneous throughput limit at the bottleneck channel based on the actual net width, and apply a social force repulsion model to individuals facing each other. Individuals who reach the safe exit are marked as having successfully evacuated and removed from the scene.

10. The method for processing simulation data on flood evacuation in riverside mountainous cities as described in claim 1, characterized in that, Differentiated spatial targeting is implemented based on heterogeneous user groups, specifically including: The population in the study area was divided into two categories: the mobile population and the resident population. The mobile population is deployed using a uniform distribution algorithm, and the early warning response delay follows a normal distribution, with both the mean and standard deviation taking larger values. The permanent residents are deployed using a random distribution algorithm weighted by the residential function of the building. The early warning response delay follows a normal distribution, with the mean and standard deviation being the smallest values.

Images