A surface runoff source-sink analysis method for urban three-dimensional landscape

Abstract

The application discloses a kind of surface runoff source and sink analysis methods for urban three-dimensional landscape, it is related to urban rain flood management and landscape planning field, including: obtaining point cloud data, meteorological hydrology data and geospatial data;From geospatial data, extract vegetation contour data and building contour data;From point cloud data, extract vegetation point cloud data and building point cloud data respectively;Calculate vegetation three-dimensional index and building three-dimensional index;According to meteorological hydrology data, rainfall runoff simulation analysis is carried out using hydrological model, obtain surface runoff depth data, calculate source and sink area ratio and its turning point;The value of vegetation three-dimensional index and building three-dimensional index at the time of simulated analysis obtained surface runoff source and sink landscape turning point is taken as the key three-dimensional index of surface runoff response.Correctly obtain and analyze the problem that existing technology is difficult to acquire and analyze urban three-dimensional landscape information, and the application improves the regulation ability to surface runoff.

Description

Technical Field

[0001] This application relates to the field of urban stormwater management and landscape planning, and more specifically, to a method for analyzing the source and sink of surface runoff for urban three-dimensional landscapes. Background Technology

[0002] Urban surface runoff is a major factor contributing to stormwater risk, resulting from extensive impermeable surfaces (such as sidewalks and buildings) created by urban development. This prevents rainwater from infiltrating, causing it to accumulate in rivers or depressions, leading to localized flooding or river overflows. With climate change causing more extreme rainfall, excessive surface runoff threatens urban infrastructure and property, posing a significant urban stormwater risk. Vegetated landscapes play a crucial role in reducing surface runoff, including urban parks, community green spaces, street trees, urban forests, and wetlands. Through ecological functions such as interception, water storage, evaporation, and infiltration, these vegetated landscapes help regulate surface runoff, promoting the restoration or reconstruction of natural hydrological cycles in cities, thereby reducing urban stormwater risk.

[0003] Currently, research on the relationship between urban landscape composition and configuration and surface runoff response mainly focuses on two-dimensional landscape indicators, such as the density, perimeter, area, fragmentation, and connectivity of landscape patches. However, these indicators rarely consider the heterogeneity of urban three-dimensional space, especially lacking detailed characterization of the vertical features of different landscape types. For example, the canopy characteristics of vegetation landscapes (leaf area, canopy coverage, canopy height) determine their runoff regulation function and services, while the roof type and volume characteristics of architectural landscapes affect runoff generation and spatial distribution. With the widespread application of LiDAR technology, the reconstruction and indicator research of urban three-dimensional landscapes have developed rapidly. However, surface runoff research based on three-dimensional landscapes still suffers from weak universality, low spatial accuracy, and insufficient consideration of the dynamics of rainfall-runoff processes. Therefore, more refined urban landscape pattern information is needed to simulate rainfall-runoff processes in urban areas, providing a more accurate basis for scientifically optimizing urban landscape composition and configuration and mitigating stormwater risks.

[0004] In related technologies, landscape configuration based on land use is a common surface runoff regulation technique. For example, Chinese patent document CN109886476A proposes a spatial model of urban hydrological processes and a land use method based on it, which improves surface runoff regulation capacity by configuring land use types of local rainwater inundation points and catchment corridors; CN115248838A discloses a method and system for optimizing the spatial layout of rainwater harvesting systems, which optimizes the spatial layout of surface runoff harvesting systems using catchment areas as units. However, the above methods are difficult to obtain surface runoff simulation results at different time steps at a fine scale, making it impossible to accurately quantify the impact of different types of landscapes on runoff volume and its flow path. Summary of the Invention

[0005] 1. Technical problems to be solved

[0006] To address the problem that existing technologies struggle to accurately acquire and analyze urban 3D landscape information, this application provides a method for analyzing the source and sink of surface runoff in urban 3D landscapes. This method can accurately extract and quantify the 3D parameters of urban vegetation and buildings by combining point cloud data obtained from airborne lidar. It can also apply hydrological models to establish scenario simulations and analyze the source-sink dynamic changes of vegetation and buildings during surface runoff. This allows for the effective design and planning of urban green infrastructure, improving its ability to regulate surface runoff and thus alleviating urban flooding.

[0007] 2. Technical Solution

[0008] The purpose of this application is achieved through the following technical solution.

[0009] A method for analyzing surface runoff sources and sinks in urban 3D landscapes includes: acquiring point cloud data, meteorological and hydrological data, and geospatial data, and preprocessing the acquired data; extracting vegetation contour data and building contour data from the preprocessed geospatial data; extracting vegetation point cloud data and building point cloud data from the preprocessed point cloud data based on the extracted vegetation and building contour data; calculating vegetation 3D indicators and building 3D indicators based on the extracted vegetation and building point cloud data; performing rainfall-runoff simulation analysis using a hydrological model based on the preprocessed meteorological and hydrological data, calculating the source-sink landscape transition process and its turning points; and calculating the values ​​of vegetation 3D indicators and building 3D indicators at the moment of the surface runoff source-sink landscape turning point obtained from the simulation analysis, which are used as key 3D indicators of surface runoff response.

[0010] The process includes: acquiring point cloud data: using airborne lidar for aerial surveying of the study area to obtain raw point cloud data containing 3D information on surface features such as buildings and vegetation. Acquiring meteorological and hydrological data: obtaining rainfall and evaporation data for different time periods from meteorological stations; obtaining hourly river flow data from hydrological stations. Acquiring geospatial data: acquiring geospatial data such as remote sensing imagery, digital elevation models, and land use data for the study area. Data preprocessing: filtering and registration of the point cloud data; quality control of meteorological data; geometric and radiometric calibration correction of geospatial data. Extracting vegetation and building outlines: extracting vegetation-covered areas and building outlines as spatial extents from the preprocessed geospatial imagery. Extracting vegetation and building point clouds: cropping corresponding vegetation and building point clouds from the preprocessed point cloud based on the extracted outline extents. Calculating 3D indices: analyzing the point cloud data to calculate 3D indices such as vegetation canopy parameters and building volume fraction. Hydrological simulation: establishing a distributed hydrological model to simulate runoff under different rainfall scenarios and obtain surface runoff depth. Identify key three-dimensional indicators: Analyze the correlation between three-dimensional indicators and runoff response indicators, and determine the values ​​of vegetation and building parameters at the inflection point of surface runoff source-sink landscape obtained from simulation analysis, which will serve as key three-dimensional indicators of surface runoff response.

[0011] Surface runoff depth data is obtained through simulation analysis of preprocessed meteorological and hydrological data using hydrological models. This data reflects the surface runoff depth of a specific region or watershed within a given period, i.e., the runoff depth formed by rainwater on the surface. A hydrological model is a mathematical tool that predicts the impact of hydrological events by simulating rainfall, snowmelt, and other meteorological and hydrological processes, as well as the interaction between surface and groundwater flows. In this case, the hydrological model is used to analyze the rainfall-runoff process. Preprocessed meteorological and hydrological data includes information such as rainfall. The hydrological model uses this data to simulate the rainfall-runoff process, i.e., how rainwater forms runoff on the surface. The simulation analysis results include surface runoff depth data, i.e., the depth of rainwater formed on the surface at the simulated time and location. This can be understood as the water depth on the surface when rainwater occurs, a quantitative data describing the accumulation and flow of rainwater.

[0012] Furthermore, vegetation point cloud data is extracted from the preprocessed point cloud data, including: using the preprocessed point cloud data as input data; obtaining point cloud data that overlaps with the spatial distribution of the vegetation outline from the input data as the original vegetation point cloud data; calculating the height, area, and volume of the original vegetation point cloud data to obtain the three-dimensional structure data of the vegetation point cloud; and calculating leaf area index, canopy height, canopy coverage, canopy height heterogeneity, and green space parameters based on the three-dimensional structure data of the vegetation point cloud as three-dimensional vegetation indicators.

[0013] The process involves several steps: Input data: Preprocessed point cloud data is used as the raw input data for vegetation point cloud extraction. Acquiring vegetation point clouds: Point cloud data that overlaps with the pre-obtained vegetation outline spatial distribution is extracted from the input raw point cloud data and used as the raw data for the vegetation point cloud. Calculating 3D structural data: Statistical analysis is performed on the extracted raw vegetation point cloud data to calculate the height, area, and volume information of the point cloud, which serves as the 3D structural data of the vegetation point cloud. Calculating vegetation indices: Based on the 3D structural data of the vegetation point cloud, parameters such as leaf area index, canopy height, canopy coverage, and canopy height heterogeneity, as well as green space-related parameters, are further calculated as 3D indices of the vegetation. Output results: Through the above process, point cloud data expressing the 3D structural characteristics of vegetation and related 3D indices are obtained, completing the extraction from raw point cloud to vegetation point cloud and vegetation 3D indices.

[0014] Leaf area index (LAI) is the ratio of leaf surface area per unit projected area of ​​the ground to the total projected area of ​​the ground. It is a key parameter for vegetation, used to quantify the density of vegetation cover and reflecting the leaf area. The LAI is obtained by calculating the surface area of ​​leaves in vegetation point cloud data, providing quantitative information about vegetation cover density. Canopy height refers to the vertical height of the upper structures of vegetation (such as trees and plants) from the ground. It reflects the vertical distribution range of vegetation. Canopy height can be measured by analyzing vegetation point cloud data, providing information about the vertical structure of vegetation. Canopy cover is the percentage of the ground surface covered by vegetation in the horizontal direction. It describes the distribution density of vegetation on the ground. Canopy cover can be calculated from vegetation point cloud data, providing information about the horizontal distribution of vegetation. Canopy height heterogeneity indicates the degree of variation in height within the vegetation canopy. Higher heterogeneity indicates greater differences in height within the vegetation. Canopy height heterogeneity can be calculated by analyzing the height distribution of vegetation point cloud data, providing information about the internal structure of the vegetation canopy. Green space parameters are parameters that describe the state and characteristics of vegetation, and can include information such as the health status and growth status of the vegetation. Based on the 3D structure data of vegetation point clouds, green space parameters can be calculated, providing more comprehensive information about the characteristics and state of the vegetation.

[0015] The technical solution for calculating the raw vegetation point cloud data includes: dividing the entire point cloud data into different categories, one of which corresponds to vegetation point clouds. This can be achieved by classifying point clouds using land use type and vegetation cover boundary. Within the vegetation point cloud categories, vegetation outlines are further extracted. This can be achieved by analyzing the shape and density of the point clouds or using morphological operations. Morphological operations can be used to detect and extract vegetation boundary points. The spatial distribution of the vegetation outline range is matched with the entire point cloud data to obtain point cloud data that overlaps with the spatial distribution of the vegetation outline range. This can be done through spatial geometric relationships and point cloud attributes, such as spatial coordinates, point cloud density, or color. Based on the matched point cloud data, the point clouds intersecting with the vegetation outline range are extracted as the raw vegetation point cloud data. This step ensures that the selected point cloud data truly belongs to the vegetation area. The raw vegetation point cloud data is further analyzed to calculate the three-dimensional structural data of the vegetation, including indicators such as height, area, and volume. This involves mathematical and geometric processing of the point cloud data, such as calculating the surface area and volume of the point cloud, and extracting vegetation height information. Based on the three-dimensional structural data of vegetation point clouds, specific vegetation indicators, such as leaf area index, canopy height, and canopy coverage, are calculated. These indicators provide detailed information about vegetation characteristics and structure.

[0016] The calculation of the 3D structure data of vegetation point clouds includes: calculating the height of the vegetation point cloud by extracting the elevation information of each point. This can be achieved using the Z-coordinate information from point cloud data collected by LiDAR or other sensors. Using the elevation information of each vegetation point, its height relative to a reference plane is calculated. Height calculation can be performed using point cloud processing libraries (such as LiDR, Open3D, etc.) or custom algorithms. For the area calculation of the vegetation point cloud, surface reconstruction or direct area estimation of the point cloud can be used. Surface reconstruction methods include triangular mesh generation, which forms a surface by connecting points in the vegetation point cloud to create triangular faces. Area calculation can be achieved by calculating the total surface area of ​​these triangles. Volume calculation of the vegetation point cloud typically requires voxelization or volume reconstruction. This is done by converting the vegetation point cloud into a 3D voxel mesh, calculating the volume of each voxel, and finally summing the results to obtain the overall vegetation volume. Point cloud-based volume reconstruction algorithms, such as Moving Least Squares (MLS), can also be used. Statistical analysis is performed on the calculated height, area, and volume data to obtain key statistical indicators such as mean and variance. Statistical methods, such as calculating the mean, standard deviation, and median, are used to understand the overall characteristics of the vegetation point cloud data. This can be achieved using point cloud processing libraries or statistical analysis software. The calculated three-dimensional structural data is then visualized to more intuitively understand the spatial distribution and characteristics of vegetation. Point cloud visualization tools, such as LiDR, Open3D, or other specialized Geographic Information System (GIS) software, are used to present the height, area, and volume information of the vegetation point cloud.

[0017] Furthermore, building point cloud data is extracted from the preprocessed point cloud data, including: using the preprocessed point cloud data as input data; extracting point cloud data that overlaps with the spatial distribution of the building outline from the input data as the original building point cloud data; calculating the height, area, perimeter, and volume of the original building point cloud data to obtain the three-dimensional structure data of the building point cloud; and calculating the roof height, slope, volume ratio, and crowding degree of the building point cloud based on the three-dimensional structure data of the building point cloud as three-dimensional indicators of the building.

[0018] The process involves several steps: Input data: Preprocessed point cloud data is used as the raw input data for building point cloud extraction. Building point cloud acquisition: Point cloud data overlapping with the pre-obtained building outline spatial distribution is extracted from the input raw point cloud data and used as the raw data for the building point cloud. 3D structure data calculation: Statistical analysis is performed on the extracted raw building point cloud data to calculate the height, area, perimeter, and volume of the point cloud, which serves as the 3D structure data of the building point cloud. Building index calculation: Based on the 3D structure data of the building point cloud, parameters such as roof height, roof slope, floor area ratio, and crowding level are further calculated as 3D indicators of the building. Output results: Through the above process, point cloud data expressing the 3D structural characteristics of the building and related 3D indicators are obtained, completing the extraction from raw point cloud to building point cloud and 3D building indicators.

[0019] Roof height refers to the vertical distance from the top of a building to the ground. It provides information about the building's vertical structure and reflects its height characteristics. The roof height can be obtained by calculating the height of the building's point cloud data, providing information about the building's vertical structure. Slope indicates the degree of inclination of the building's roof surface relative to the horizontal plane. A larger slope indicates a higher roof inclination. The roof slope can be calculated by analyzing the building point cloud data, providing information about the roof surface inclination. Floor area ratio (FAR) is the ratio of the total building area to the land area. It reflects the density of buildings on the ground and the extent of their land occupation. The FAR can be obtained by calculating the volume and land area of ​​the building point cloud data, providing information about the building's utilization density on the ground. Crowding refers to the degree of compactness of buildings in space, i.e., the relative density between buildings. A larger crowding indicates a denser building distribution. The crowding can be calculated by analyzing the distribution of the building point cloud data, providing information about the relative positional relationships between buildings. The technical solution for obtaining the raw data of the building point cloud is the same as that for the raw data of the vegetation point cloud, and will not be repeated here. The technical solutions for calculating the three-dimensional structural data of building point clouds are the same as those for calculating the three-dimensional structural data of vegetation point clouds, and will not be described in detail here.

[0020] Furthermore, the acquired meteorological and hydrological data are preprocessed, including: determining different rainfall events by rainfall intervals, and acquiring hourly rainfall data and hourly flow data. Hourly rainfall data represents the amount of rainfall per unit time, and hourly flow data represents the outflow of the catchment area per unit time.

[0021] The process involves the following steps: First, acquiring raw data: obtaining raw hourly rainfall data from meteorological stations and raw hourly flow data from hydrological stations. Second, determining rainfall events: dividing the observation period into different rainfall events based on the intervals between rainfall events. Third, extracting hourly rainfall data: extracting the corresponding hourly rainfall data for each determined rainfall event, i.e., the rainfall value for each time interval. Fourth, extracting hourly flow data: extracting the corresponding hourly flow data for each determined rainfall event, i.e., the river flow value for each time interval. Fifth, recording data characteristics: identifying hourly rainfall data as representing the rainfall amount per unit time, and hourly flow data as representing the outflow from the catchment area per unit time. Sixth, outputting preprocessing results: after the above process, the raw meteorological and hydrological data are preprocessed into hourly rainfall and flow data corresponding to the rainfall events.

[0022] Furthermore, the acquired geospatial data is preprocessed, including: projecting the acquired geospatial data using the Albers equal-area conic projection to generate vector map data; registering the acquired geospatial data with known control points to generate registered image data; and associating the generated vector map data with the registered image data as preprocessed geospatial data.

[0023] The process includes the following steps: Projection transformation: The original geospatial data is transformed using the Albers equal-area conic projection to generate vector map data in a planar coordinate system. Georegistration: Using known control points, the acquired original geospatial image data and map data are registered in a geographic coordinate system to generate registered image data. Association and organization: The transformed vector map data and the registered image data are associated and organized to establish a spatial correspondence between them. Output preprocessing results: Through the above transformation, registration, and association processes, the geospatial data is transformed from its raw state to a preprocessed result that meets the analysis requirements.

[0024] The Albers Equal Area Conic Projection is a coordinate transformation method for conic projections. The projection surface is a combination of two mutually tangent cones, coplanar and symmetrically distributed on either side of the Earth's poles. The ellipse generated by the intersection of the cones is distortion-free, ensuring that the projected area equals the corresponding actual ground area, achieving equal-area mapping. This projection method can significantly reduce deformation in mid-latitude regions, making it suitable for mapping mid-latitude areas. It is commonly used for map production in areas where area calculations are sensitive, such as calculating land use area.

[0025] Geospatial registration is a method for correcting geospatial data, aiming to achieve optimal matching of geographic data from different sources within a geographic coordinate system. The process involves collecting ground control points with known coordinates, such as those of identifiable objects like traffic intersections and river confluences. The corresponding positions of these control points are then marked on the geospatial image. A geometric transformation algorithm is applied to align and register the image points with the geographic control points. This process is iterated until the error between the image and the control points is less than the tolerance. Finally, a geospatial image aligned with the standard geographic coordinate system is obtained. Geospatial registration corrects offsets between different images, achieving precise correspondence with the geographic coordinate system and making image analysis results more reliable for spatial reference.

[0026] Furthermore, rainfall-runoff simulation analysis is conducted using hydrological models, including: constructing a distributed hydrological model with a grid size of N, which includes surface runoff sub-models, saturated zone sub-models, unsaturated zone models, and river channel sub-models; and conducting rainfall-runoff simulation analysis using the hydrological model, including: calculating rainfall amounts at different return periods using the storm intensity formula, the basic form of which is shown below:

[0027]

[0028] Represents rainfall intensity in mm / min. Represents the design return period for rainfall. The duration of the designed rainfall (min) is represented by the parameters A1, C, n, and b, which are determined according to the rainfall intensity formula for different regions. Based on the calculated rainfall amounts for different return periods, a rainfall process line corresponding to the return period is set as the rainfall input scenario, where the rainfall process line represents the rainfall intensity-time distribution.

[0029] The model comprises the following components: Surface runoff sub-model: This model simulates surface runoff processes, calculating runoff entering the river network. Saturated zone sub-model: This model simulates water movement in near-saturated soil zones, following linear seepage equations. Unsaturated zone model: This model simulates water movement in unsaturated soil zones, following Richard's equations. River channel sub-model: This model simulates hydrodynamic processes in rivers and lakes, calculating flow and water level changes. Distributed hydrological model: This framework consists of multiple sub-models coupled through the exchange of simulation results, enabling the simulation of distributed spatial hydrological processes. In summary, these sub-models collectively constitute a complete distributed hydrological model, used to simulate hydrological processes in different components, ultimately achieving the goal of simulating the rainfall-runoff process of the entire watershed.

[0030] The rainfall input scenario is the rainfall intensity-time distribution data set during the hydrological simulation process. Based on the rainfall characteristics of the study area, design storms with different return periods are selected, and the rainfall amounts under different return periods are calculated. A rainfall intensity-time distribution curve is generated based on the calculated rainfall amounts, describing the temporal progression of the rainfall event; this is the rainfall hydrograph. The rainfall hydrographs with different return periods are pre-set as input files for the hydrological model and called when running the model; this is the rainfall input scenario. Setting multiple rainfall input scenarios with different return periods allows the hydrological model to simulate under various conditions. In summary, the rainfall input scenario is the rainfall intensity-time distribution data used to drive hydrological simulation; by setting rainfall input scenarios with different return periods, rainfall-runoff processes under different conditions can be simulated.

[0031] Furthermore, rainfall-runoff simulation analysis is conducted using a hydrological model, including: taking grid data from the surface runoff sub-model and the river sub-model as input, and using a spatial overlap algorithm to extract the spatial intersection grids of the surface runoff sub-model and the river sub-model as water exchange connection units; establishing river water balance equations and surface water balance equations on the water exchange connection units based on the coupling relationship between the surface runoff process and the river runoff process; using the established river water balance equations and surface water balance equations as input to the equation set, and employing the Saint-Venant equations numerical calculation method and the finite element discretization method to perform numerical calculations and grid discretization on the equation set, obtaining the coupled calculation results of the river hydrodynamic model and the surface hydrological model, which serve as the hydrological model.

[0032] The spatial overlap algorithm is used to determine whether two spatial grids or layers intersect in space. It takes two spatial grid layers as input and uses spatial relationship operations to determine if the geometric extents of each grid intersect. If the geometric extents of the two grids intersect, the new grid formed by the intersection is extracted. The final output is the grid elements that intersect in space. This algorithm can determine the topological relationship between two spatial grids and identify grid cells that overlap in a certain area. In the coupling of hydrological models, it can be used to extract the junction cells of river channels and surface grids, realizing the technical implementation of water exchange.

[0033] Within each river channel calculation unit, a river channel water balance equation is established based on the balance relationship between river inflow and outflow: Q out =Q in +Q lateral -Q overflow +PE, where Qin is the river inflow and Q out Q represents the river's outflow. lateral For the bifurcation flow rate, Q overflowLet P be the overflow, P be the rainfall, and E be the evapotranspiration. In each surface calculation unit, based on the balance relationship between surface water inflow and outflow, a surface water balance equation is established: Q out =Q in +PEI, where Qin is the surface water inflow, Q out Let P be the surface outflow, E be the rainfall, E be the evaporation, and I be the surface infiltration. In the handover computational unit, the two equations are coupled by exchanging the water volume between the river channel and the surface to achieve coupled simulation.

[0034] The Saint-Venant equations are a set of differential equations describing fluid motion, jointly derived by the French mathematician Claude Louis Marie Henri Navier and the British mathematical physicist George Gabriel Stokes. Its basic components include: the continuity equation (describing the conservation of fluid mass); the equation of motion (describing changes in fluid motion); and the energy equation (describing the fluid's energy balance). The variables in these three equations include velocity, pressure, and density, and the three equations are coupled together to describe the flow patterns of fluids. The Saint-Venant equations are the most fundamental set of equations describing fluid motion and are widely used in computational fluid dynamics, hydrology, and other fields, enabling the creation of accurate physical motion simulations.

[0035] The establishment of the hydrological model includes: using the established river channel water balance equations and surface water balance equations as inputs to the Saint-Venant equations; employing the Saint-Venant finite element discretization scheme to discretize the river channel and surface equations in both the time and spatial domains, transforming them into a system of algebraic equations; and using numerical calculation methods such as Newton's iteration method to solve the discretized equations, obtaining the water volume and flow rate for each computational unit at each time period. The coupling of the river channel model and the surface model is achieved through water volume exchange between computational units. The above calculation steps are repeated to simulate the entire computational time domain, obtaining the hydrodynamic process results of the coupled river channel and surface model. Post-processing of the calculation results, including verification and visualization, is performed to generate the final simulation result output. This completes the establishment of a coupled hydrological process model based on numerical calculation and finite element discretization of the governing equations.

[0036] Furthermore, rainfall-runoff simulation analysis is conducted using a hydrological model, including: inputting rainfall amounts with different return periods as input data into the established hydrological model; calibrating the hydrological model using hourly rainfall and hourly flow data as calibration data; obtaining the sensitivity indicators of the calibrated hydrological model using a global sensitivity algorithm; obtaining the value combinations of the sensitivity indicators using the Monte Carlo sampling method, inputting the value combinations into the calibrated hydrological model for simulation calculation, and obtaining the model output results; comparing the statistical characteristic indicators of the model output results with those of the calibration data, selecting the sensitivity indicators corresponding to the value combinations of the statistical characteristic indicators that meet the threshold, and using the selected sensitivity indicators as the optimal parameters of the hydrological model.

[0037] Hourly rainfall data refers to the rainfall recorded every hour. This data typically represents the intensity, distribution, and spatiotemporal variation of precipitation within a region over a certain time period. Hourly rainfall data is used to simulate precipitation processes in the calibration of hydrological models. By comparing it with measured data, the parameters of the hydrological model can be adjusted to more accurately simulate actual rainfall input. Hourly flow data refers to the flow rate recorded every hour. This data represents the change in water volume in a catchment area over time. Hourly flow data is used to verify the effectiveness of the hydrological model in simulating water flow processes. By comparing it with measured flow data, the performance of the hydrological model can be evaluated and adjustments made to improve the accuracy of the simulation results. The combined use of hourly rainfall and hourly flow data during hydrological model calibration improves the model's reliability and adaptability. By comparing simulation results with measured data, the parameters of the hydrological model can be adjusted to better reflect actual hydrological processes.

[0038] The global sensitivity algorithm is a method for sensitivity analysis of model parameters. It systematically varies the model parameters, covering their entire range. The changes in model output are recorded accordingly. By statistically analyzing the relationship between the range of parameter variations and the range of model output variations, the global sensitivity of each parameter to the model results is calculated. This analysis is performed on all parameters to obtain a global sensitivity index for each parameter. The importance of each parameter is then ranked based on its sensitivity index. Compared to local sensitivity analysis, global sensitivity analysis can more comprehensively reflect the impact of parameters on model results and is used for model calibration and identification of key parameters.

[0039] Sensitive indicators are parameters or variables in a hydrological model that are sensitive to simulation results. These can include specific variables in the model output, such as peak flood discharge or runoff. By defining sensitive indicators, we can focus on the most critical variables in the model output, allowing for more effective exploration of the parameter space in Monte Carlo sampling. Monte Carlo sampling is a random sampling method that obtains multiple sets of parameter values ​​by randomly sampling within the parameter space. Monte Carlo sampling generates a large number of parameter combinations, covering a wide range of the parameter space, to evaluate the model's performance under different parameter combinations. The parameter combinations obtained from Monte Carlo sampling are input into a calibrated hydrological model for simulation calculations, yielding the corresponding model output results. Through simulation calculations, we can obtain the hydrological model's response for each set of parameter values, including the values ​​of the sensitive indicators in the model output. This forms a set of model outputs in the parameter space. Analyzing the model output results evaluates the model's performance under different parameter combinations, paying particular attention to changes in sensitive indicators. By analyzing the model output results obtained from Monte Carlo sampling, we can determine which parameters have a significant impact on the sensitive indicators in the model output. This helps in understanding model uncertainties and improving model reliability.

[0040] The calculation of optimal parameters includes: statistical characteristic indicators, which encompass various statistical properties between the model output and calibration data, such as root mean square error (RMSE), correlation coefficient (R²), and Nash efficiency coefficient (NSE). These indicators measure the model's predictive performance and determine the goodness of fit of the model output by comparing it with the calibration data. After performing model simulation calculations for each set of parameter values, the statistical characteristic indicators of the model output and calibration data are calculated. By comparing the statistical characteristic indicators, the model's performance under different parameter combinations can be evaluated, and which combinations can better fit the calibration data can be identified. Thresholds for the statistical characteristic indicators are set as criteria for judging the model's performance. By setting thresholds, parameter combinations that perform well in the model output can be selected; combinations that meet the thresholds are considered acceptable. Sensitive indicators are selected from the parameter combinations that meet the thresholds. These sensitive indicators are key variables in the model output and are of great significance for describing hydrological processes. Selecting sensitive indicators helps to more accurately understand the model's simulation effect on hydrological processes, so as to further determine the optimal parameters. The selected sensitive indicators are used as the optimal parameters of the hydrological model, forming the optimal parameter combination. The optimal parameter combination is the best solution obtained by calibrating the model while meeting the threshold values ​​of statistical characteristic indicators. These parameters enable the model to better fit actual hydrological data.

[0041] Furthermore, rainfall-runoff simulation analysis is conducted using hydrological models, including: taking rainfall input scenarios corresponding to different return periods as input and importing them into a hydrological model with determined optimal parameters; for each rainfall input scenario, calculating surface runoff depth data under different rainfall steps using the hydrological model with determined optimal parameters; generating surface runoff distribution maps for corresponding return periods using the calculated surface runoff depth data under different rainfall steps; and dividing green infrastructure into source areas and sink areas using the surface runoff distribution maps, where source areas represent units that generate runoff and sink areas represent units that do not generate runoff.

[0042] The area ratio of the source and sink regions was statistically calculated, and a polynomial model was established based on the rainfall duration and the ratio of source and sink areas. A threshold regression algorithm was applied to calculate the turning point in the dynamic change process of the source region based on the established polynomial model. The turning point represents the point where the slope of the ratio of sink to source region changes abruptly. At the turning point in the dynamic change process of the source and sink regions, the corresponding three-dimensional vegetation index and three-dimensional building index were obtained as key three-dimensional indicators of surface runoff response.

[0043] The calculation of surface runoff depth data under different rainfall steps includes: setting rainfall inputs with different time steps, such as 10 minutes, 15 minutes, and 1 hour. In the hydrological model, a calculation time step is set, for example, 1 hour. For each rainfall input scenario, input files with different rainfall steps are aggregated based on the model's calculation step. The hydrological model reads these input files and runs the simulation. The model iteratively calculates the surface runoff depth at each calculation step and outputs the surface runoff depth for each time period. The surface runoff depth results under different rainfall steps are compared and analyzed. The impact of different input rainfall steps on the simulation results is evaluated, providing a basis for subsequent model optimization.

[0044] The process of generating surface runoff distribution maps includes: collecting DEM data of the study area and extracting the watershed extent; establishing a grid for the study area on a GIS platform; importing surface runoff depth results from different grid cells calculated in the hydrological model into the GIS platform and spatially connecting them with the grid layers; interpolating runoff depth of grid cells using the spatial analysis module based on simulation results for different return periods; cropping the watershed extent to obtain surface runoff distribution maps of the study watershed under different return periods; setting color gradients corresponding to runoff depths to generate color-coded visualizations of surface runoff distribution; and outputting surface runoff distribution maps under different return period conditions.

[0045] The process of dividing green infrastructure into source and sink areas includes: obtaining a spatial distribution map of surface runoff, which can be obtained through remote sensing data, Geographic Information System (GIS) data, or hydrological model simulations. This image should reflect the spatial variation of surface runoff within the region. A threshold is set to distinguish between areas that generate runoff and those that do not. This threshold can be based on the water depth of depressions, for example, setting a surface runoff depth greater than the water depth of the depression. Image processing techniques are used to segment and classify the surface runoff distribution map. This can be achieved through thresholding, clustering algorithms, or deep learning methods. The goal is to divide the surface runoff image into source and sink areas. Source areas, i.e., areas determined to generate surface runoff, are extracted from the classified image. This involves extracting categories or pixels in the image that are associated with source areas. Similarly, sink areas, i.e., areas that do not generate surface runoff, are extracted from the classified image. This is achieved by excluding categories or pixels that are not associated with source areas. The extracted source and sink areas are validated using field survey data or other reference data. If necessary, thresholds or algorithm parameters can be adjusted to improve the accuracy of the results. The results can be imported into a GIS for visualization, analysis, and further spatial data processing. GIS tools provide a deeper analysis and management of surface runoff distribution. Combining this with other geographic information, such as topography and land use data, can further refine the spatial characteristics of source and sink areas, thereby improving the accuracy of delineation.

[0046] The process involves using a DEM (Digital Elevation Model) to extract the watershed extent, calculating the flow direction at each grid point, and determining the source and sink areas. The areas of the source and sink areas are statistically analyzed at different time periods, and their area ratios are calculated. A multinomial regression model is established with rainfall duration as the independent variable and the source-sink area ratio as the dependent variable. A threshold regression model is applied to calculate the inflection point of the source-sink area ratio; for example, if there exists a point where the slope of the fitted model before that point is negative and the slope of the fitted model after that point is positive, then that point is the inflection point. The slope of the fitted model of the source-sink area ratio and rainfall duration is calculated on the multinomial model curve, and abrupt changes in the slope are identified. This abrupt change point is the inflection point where the source area transforms into the sink area. Repeating the above process allows for the calculation of inflection points under different rainfall scenarios.

[0047] The key three-dimensional indicators for calculating surface runoff response include: collecting three-dimensional data on vegetation cover and buildings within the extracted source and sink regions; calculating vegetation three-dimensional indicators, such as leaf area index and canopy volume; calculating building three-dimensional indicators, such as building density and building height; extracting vegetation and building indices at the calculated source-sink inflection points; and statistically analyzing the distribution of index values ​​at the inflection points. The final vegetation and building index values ​​can represent the key three-dimensional indicators of surface runoff response.

[0048] Furthermore, the key three-dimensional indicators for calculating surface runoff response include: using simulated surface runoff depth data to calculate surface runoff response indicators, which include runoff volume and runoff length index; calculating the correlation coefficient between surface runoff response indicators and vegetation three-dimensional indicators to obtain vegetation landscape correlation coefficients; calculating the correlation coefficient between surface runoff response indicators and building three-dimensional indicators to obtain building landscape correlation coefficients; at the turning points in the dynamic changes of the source and sink areas, obtaining the vegetation landscape correlation coefficients and building landscape correlation coefficients corresponding to the turning points as three-dimensional landscape indices; and using the obtained three-dimensional landscape indices as key three-dimensional indicators for surface runoff response.

[0049] Runoff volume refers to the total amount of water flowing through the surface within a certain period. It represents the total amount of runoff caused by hydrological processes such as rainwater or snowmelt, and is usually expressed in volume units (such as cubic meters or cubic feet). Runoff volume is one of the main indicators for assessing the runoff generated by water bodies during hydrological processes. Through simulation calculations, runoff volume at different time periods can be obtained, helping to understand the distribution and changes of rainwater within a watershed. The flow length index describes the path of water flowing across the surface. It represents the average path length of water flowing through the surface. The larger the flow length index, the longer the path the water body travels during surface runoff. The flow length index provides information about the runoff path, helping to understand how water moves within a watershed. A larger flow length index indicates that the water body has experienced a relatively long path during surface flow, while a smaller flow length index indicates a shorter runoff path.

[0050] The calculation of vegetation-landscape correlation coefficients includes: using vegetation point cloud computing to calculate vegetation structure parameters of the study area, such as leaf area index, green volume, canopy height, canopy coverage, and vegetation height heterogeneity; using hydrological models to simulate and calculate surface runoff response indicators for corresponding time phases, such as runoff volume and runoff length index; matching vegetation index and runoff response indicator data for different plot units; and calculating the Pearson correlation coefficient between vegetation indices and runoff response indicators. The vegetation-landscape correlation coefficient can be used to assess the impact of vegetation on surface runoff in different regions. The calculation of building-landscape correlation coefficients is the same as that of vegetation-landscape correlation coefficients and will not be repeated here.

[0051] In this process, at the extracted source-sink transition points, vegetation cover and building data are determined at the corresponding transition times. Various vegetation indices, such as leaf area index and canopy cover, are calculated at these transition times. Using pre-established vegetation-landscape correlation coefficients, key three-dimensional vegetation landscape indices are extracted for each transition time. Similarly, various building indices are calculated at the transition times, and key building three-dimensional landscape indices are extracted using building-landscape correlation coefficients. The vegetation and building landscape indices are combined to form a three-dimensional landscape index. This process is repeated to obtain the corresponding three-dimensional landscape indices at different transition points. Finally, the key three-dimensional landscape indices for the time series during the source-sink transition process are obtained. This index reflects the comprehensive impact of landscape elements on runoff response.

[0052] Compared to existing technologies, the advantages of this application are:

[0053] (1) Obtain the original point cloud data, which contains the three-dimensional coordinate information of various urban features. Remove outliers and noise points through preprocessing to ensure data quality. Obtain geospatial image data, which contains urban landscape planar range information, and perform projection conversion, registration and other processing to accurately correspond to the point cloud data. Extract the vegetation outline and building outline from the geospatial data as parameters to calculate the spatial range. Based on the extracted outline range, accurately crop the point cloud data to obtain vegetation point cloud and building point cloud. Analyze the point cloud data to calculate various three-dimensional parameters of vegetation and buildings, such as canopy height and volume ratio. The accurate acquisition and quantification of the three-dimensional structural features of the main urban features are realized, overcoming the problem of poor data quality caused by the reliance on subjective judgment in the existing technology, making the three-dimensional landscape analysis of the city more accurate.

[0054] (2) Construct a distributed hydrological model that includes sub-models such as surface runoff and river runoff, and set different rainfall scenarios; import the extracted parameters of vegetation and buildings, such as canopy height and building density, into the hydrological model; run the hydrological model to simulate the surface runoff process and obtain the source and sink state changes of vegetation and buildings; evaluate the runoff regulation capacity of different urban green spaces based on the source and sink state changes; fully consider the impact of dynamic changes in landscape parameters on runoff, accurately determine the source and sink function conversion of vegetation and buildings, overcome the shortcomings of static landscape assessment in existing studies, and realize the accurate simulation and assessment of the surface runoff process mechanism;

[0055] (3) Based on the hydrological model simulation, the turning point of the source area and sink area is calculated as the threshold of the surface runoff process. According to the threshold, the correlation between landscape parameters and runoff response is statistically analyzed to determine the runoff regulation capacity of different green spaces under rainfall conditions, optimize the green space layout scheme, and make the key parameters reach the threshold requirements. This provides a quantitative planning basis, which is conducive to improving the regulation role of green spaces on runoff in a targeted manner, solving the limitations of existing empirical design, and making green infrastructure planning more targeted and efficient.

[0056] This application, by combining airborne lidar point cloud data and geospatial data, can accurately extract and quantify urban vegetation and building landscape parameters containing three-dimensional structural information, thus achieving an accurate depiction of complex urban three-dimensional landscapes. By applying a spatially distributed hydrological model and inputting three-dimensional landscape parameters, it can simulate the source-sink function transformation of vegetation and buildings in the surface runoff process, thus reproducing the dynamic process of landscape source-sink. By simulating and analyzing the relationship between vegetation and building parameters and surface runoff response, it can determine the key three-dimensional landscape parameters affecting the runoff process, thus identifying the characteristics of the inflection point of the source-sink process. Attached Figure Description

[0057] Figure 1 This is a flowchart illustrating the multi-source data input and processing process of this application;

[0058] Figure 2 This is a schematic diagram of the three-dimensional landscape reconstruction process in this application;

[0059] Figure 3 This is a flowchart simulating the dynamic process of source and sink in this application;

[0060] Figure 4 This is a flowchart illustrating the process of identifying key three-dimensional landscape indicators in this application;

[0061] Figure 5 This is a schematic diagram of three-dimensional landscape indicators in one embodiment of this application;

[0062] Figure 6 This is a schematic diagram of the dynamic changes in the source-sink landscape in one embodiment of this application;

[0063] Figure 7 This is a schematic diagram illustrating the nonlinear relationship between the source-sink landscape and the rainfall time step under rainfall scenario S1 in one embodiment of this application;

[0064] Figure 8 This is a schematic diagram illustrating the nonlinear relationship between the source-sink landscape and the rainfall time step under rainfall scenario S2 in one embodiment of this application;

[0065] Figure 9 This is a schematic diagram illustrating the nonlinear relationship between the source-sink landscape and the rainfall time step under rainfall scenario S3 in one embodiment of this application;

[0066] Figure 10 This is a schematic diagram illustrating the identification of key three-dimensional landscape indicators in one embodiment of this application. Detailed Implementation

[0067] This application takes a town polder area of ​​5.13 square kilometers as an example and implements the following technical methods: Point cloud data, geospatial data, and meteorological and hydrological data of the case area are acquired and preprocessed. Parameters of vegetation and buildings are extracted as three-dimensional landscape indicators. A distributed hydrological model of the case area is constructed, and rainfall input scenarios with different return periods are set to simulate surface runoff processes. Under different rainfall scenarios, the source and sink states of vegetation and buildings at each time period are extracted, and a dynamic map of the spatial distribution of sources and sinks is drawn. The correlation between vegetation and building parameters and surface runoff response indicators is analyzed to determine the key three-dimensional indicators affecting surface runoff response. The turning point of source-sink transformation is identified, and the key three-dimensional indicators affecting surface runoff response at this turning point are determined. Through the technical implementation process of this specific embodiment, the source-sink process and key three-dimensional indicators of the case area are determined, demonstrating the technical feasibility of the proposed solution.

[0068] Figure 1 This paper presents a flowchart illustrating the multi-source data input and processing process of this application. It provides a multi-source data input and processing technology combining airborne radar point clouds and geospatial information, including the following steps: Acquiring high-resolution point cloud data from an airborne LiDAR system, including details such as urban topography and buildings. Collecting meteorological and hydrological observation data, including rainfall intensity, temperature, and humidity. Acquiring river network survey data to obtain information about the urban water system. Conducting land use and building outline surveys to obtain the spatial distribution of urban features. Registering the point cloud data to ensure consistency between point cloud data collected at different times and locations. Cropping the point cloud data to remove irrelevant information and focus on areas of interest within the city. Denoising and filtering the point cloud data to eliminate noise during acquisition and improve data quality. Based on meteorological and hydrological observation data, dividing rainfall events and determining the simulation time range and rainfall intensity. Projecting and transforming the geospatial data to the CGCS20003DegreeGKZone40 coordinate system to ensure consistency between different datasets. Performing georegistration to integrate various geospatial data into the same coordinate system. A 3D landscape model of the city is constructed using processed point cloud data. River network data is integrated into the 3D landscape to form a complete water system. Hydrological models are applied, combined with rainfall events and topographic information, to simulate surface runoff processes. The dynamic changes of vegetation and buildings in surface runoff are analyzed, and their ability to regulate runoff is assessed.

[0069] Figure 2This paper presents a flowchart illustrating the 3D landscape reconstruction process for this application. It provides a method for classifying point cloud information and calculating 3D landscape indicators, including the following steps: Utilizing point cloud data and geospatial data (vegetation cover and building outlines) obtained through the aforementioned process; performing overlap analysis between the point cloud data and the vegetation cover to extract vegetation point clouds; performing overlap analysis between the point cloud data and the building outline to extract building point clouds; performing quality control on the classified point cloud data to remove noise and outliers; using the vegetation point clouds to perform 3D reconstruction of the vegetation structure, including indicators such as leaf area index, canopy height, canopy coverage, and canopy height heterogeneity; using the building point clouds to perform 3D reconstruction of the building structure, including indicators such as roof height, roof slope, building morphology coefficient, building volume ratio, and building crowding; calculating indicators such as leaf area index, canopy height, canopy coverage, and canopy height heterogeneity for the vegetation structure; and calculating indicators such as roof height, roof slope, building morphology coefficient, building volume ratio, and building crowding for the building structure. The relevant calculation formulas are shown in Table 1.

[0070] Table 1. Calculation Methods for Three-Dimensional Landscape Indicators

[0071]

[0072] in, The area of ​​the land parcel. The floor area of ​​the building. For the number of floors, For the number of buildings, For building height, The perimeter of the building's base area. For building volume.

[0073] Figure 3 This paper presents a flowchart illustrating the dynamic process simulation of surface runoff sources and sinks in this application. It provides a hydrological simulation-based analysis technique for the dynamic process of surface runoff sources and sinks, comprising the following steps: Pre-processed measured data, including slope flow, saturated zone, unsaturated zone, and river channel. The MIKESHE / 11 model is used, which supports the coupling of modules such as slope flow, saturated zone, unsaturated zone, and river channel. The model domain is divided into a 5-meter precision grid to ensure sufficient spatial resolution. Spatial pixels where water exchange occurs between the surface and the river channel are identified, and parameters such as riverbank roughness, slope coefficient, and riverbank elevation are set. The hydrological and hydrodynamic models are coupled to ensure the simulation of water exchange between the surface and the river channel. Sensitive parameters are selected, such as vertical hydraulic conductivity, horizontal hydraulic conductivity, canopy interception ratio, depression water storage depth, and drainage time constant. Model parameters are validated using flood season data from 2022–2023. The model was calibrated using metrics such as the Nash efficiency coefficient (NSE) and correlation coefficient (R²) to ensure a good fit between the simulation results and the measured data. The specific calculation formulas are as follows:

[0074]

[0075]

[0076] and These represent the observed and simulated runoff responses, respectively. and These represent the average values ​​of the observation and simulation results, respectively. The average values ​​of the observation and simulation results are compared, and quantitative evaluation is performed using statistical indicators (such as mean absolute error, Nash efficiency coefficient, etc.). In summary, the parameter calibration results of the hydrological model are shown in Table 2 below, demonstrating the model's good fit.

[0077] Table 2. Results of Hydrological Model Calibration and Validation

[0078]

[0079] Design 24-hour duration rainstorm scenarios with different return periods (20-year and 50-year return periods), and calculate surface runoff inundation results at different time steps using these scenarios. This process helps assess extreme scenarios of urban surface runoff and provides important references for hydrological model design and urban planning. A 24-hour duration rainstorm is selected to simulate a long-term rainfall process. Rainstorms with different return periods are designed, including 20-year and 50-year return periods. Based on the designed rainstorm scenarios, surface runoff inundation calculations are performed. Rainstorm data is input into the hydrological model to simulate runoff generation at different time steps. Combined with a hydrodynamic model, surface runoff inundation at different time steps is simulated. The specific calculation formula is as follows:

[0080]

[0081] Represents rainfall intensity (mm / min). Represents the design return period for rainfall (year). The duration (in minutes) represents the design rainfall. The Chicago rain pattern distribution rainfall process line was selected, and the peak rainfall coefficient was set to 0.74 based on the conditions of the study area. The rainfall scenarios are shown in Table 3.

[0082] Table 3. Rainfall Scenario Settings

[0083]

[0084] Finally, based on the surface runoff inundation depth results, the surface landscape at each simulation step is divided into two categories: source landscape and sink landscape. The source landscape includes building units and submerged vegetation units, while the sink landscape includes unsubmerged vegetation units. For each time step, the source-sink landscape ratio, i.e., the proportion of the source landscape in the total landscape, is calculated. The time step and the source-sink ratio are fitted with a polynomial function, and an appropriate polynomial degree is selected. The correlation coefficient (R²) and the Akaike Information Content Criterion (AIC) are used to evaluate the goodness of fit, ensuring that the model is both accurate and concise in describing the data. A piecewise threshold regression model is used to find the inflection points of the source-sink ratio under different rainfall scenarios. This helps to identify key time points between source and sink landscapes under different conditions. After obtaining the fitted function and inflection points, the application of the results in urban hydrological models is analyzed. By evaluating the fitting results, the complex relationship between the time step and the source-sink ratio can be better understood, providing a deeper understanding for urban hydrological research. The polynomial function is as follows:

[0085]

[0086] This represents the independent variable, in this case, the time step of rainfall. This represents the dependent variable, in this case, the source-sink ratio. This is a constant term. The comparison of the goodness of fit between the nonlinear and linear models is shown in Table 4:

[0087] Table 4. Goodness-of-fit results for nonlinear and linear models

[0088]

[0089] The piecewise threshold regression model is as follows:

[0090]

[0091] Represents a constant. Represents the threshold parameter. Representing predictors with a threshold effect, Represents the dependent variable. Indicates additional predictor variables. Represents the hinge function, in Time equals In other cases, it equals 0.

[0092] Table 5 shows the results of identifying the inflection point of the source-sink ratio under different rainfall scenarios:

[0093] Table 5. Results of Identification of Turning Points in the Source-to-Sink Ratio

[0094]

[0095] Figure 4 This paper presents a flowchart illustrating the identification process for key three-dimensional landscape indicators in this application. The identification process, including the following steps, involves: calculating runoff response indicators at the source-sink ratio inflection point based on runoff volume and flow length index; statistically analyzing runoff response indicators and three-dimensional landscape indicators by land use unit; performing Pearson correlation analysis to determine the relationship between the three-dimensional landscape indicators and the runoff response indicators; extracting the spatial range from the source landscape to the sink landscape after the inflection point; statistically analyzing the value ranges of key three-dimensional indicators in different land use units; calculating and obtaining the mean values ​​of the above indicators as the three-dimensional indicator values ​​for different land use types; calculating runoff response indicators at the source-sink ratio inflection point based on runoff volume and flow length index; performing Pearson correlation analysis on runoff response indicators and three-dimensional landscape indicators to identify their relationship; extracting the spatial range from the source landscape to the sink landscape for further analysis; statistically analyzing the value ranges of key three-dimensional indicators in different land use units within the extracted spatial range; and calculating and obtaining the mean values ​​of the above indicators as the three-dimensional indicator thresholds for different land use types. The obtained three-dimensional indicators can be applied to the simulation and prediction of source-sink landscape dynamic processes. By evaluating the results, the applicability of these key three-dimensional indicators under different land use types can be verified, and more accurate parameter settings can be provided for urban hydrological models.

[0096] Figure 5 This is a schematic diagram of three-dimensional landscape indicators in one embodiment of this application, showing the calculation results of three-dimensional landscape indicators for vegetation and buildings in the case study area, reflecting the spatial heterogeneity of the three-dimensional landscape characteristics of the case study area. The vegetation three-dimensional landscape indicators include: Canopy Height (CH): measures the vertical height of the vegetation canopy. Canopy Coverage (CC): represents the proportion of the ground covered by vegetation. Three-dimensional Green Volume (GV): assesses the green volume of vegetation. Leaf Area Index (LAI): measures the leaf area of ​​vegetation per unit ground area. Canopy Height Heterogeneity (CH_STD): represents the degree of heterogeneity in vegetation canopy height. The building three-dimensional landscape indicators include: Building Roof Height (RH): represents the vertical height of the building roof. Building Roof Slope (RS): describes the slope of the building roof. Building Floor Area Ratio (FAR): represents the ratio of building volume to land area. Building Crowding (BCD): measures the spatial density of buildings. Building Form Factor (BSC): represents the building form index. Through these indicators, Figure 5 This demonstrates the heterogeneity of the spatial distribution of vegetation and buildings in the case study area. Different colors or shapes represent different index values, thus showcasing the diversity and spatial heterogeneity of vegetation and buildings.

[0097] Figure 6This diagram illustrates the dynamic changes in the source-sink landscape in one embodiment of this application. The source landscape initially increases and then decreases during the rainfall-runoff process, while the sink landscape exhibits the opposite trend. At the beginning of rainfall, the source landscape shows an increasing trend due to the increase in building units and submerged vegetation units caused by rainwater. As time progresses and the rainfall event develops, the source landscape gradually decreases, indicating a gradual decline in the water level of submerged vegetation units or other related factors. Conversely, the sink landscape initially decreases and then gradually increases over time. This reflects the change in the area of ​​unsubmerged vegetation units. Figure 6 The results reflect the dynamic changes in the source-sink landscape during the rainfall-runoff process. This dynamic change is of significant reference value for understanding urban hydrological processes and improving hydrological models. Specifically, S1t=1hp=0.04mm (A), S1t=12hp=9.12mm (B), S1t=24hp=2.15mm (C); S2t=1hp=3.55mm (D), S2t=12hp=33.32mm (E), S2t=24hp=3.95mm (F); S3t=1hp=4.02mm (G), S3t=12hp=37.75mm (H), S3t=24hp=4.49mm (I).

[0098] Figure 7 This is a schematic diagram illustrating the nonlinear relationship between the source-sink landscape and the rainfall time step under rainfall scenario S1 in one embodiment of this application. Figure 8 This is a schematic diagram illustrating the nonlinear relationship between the source-sink landscape and the rainfall time step under rainfall scenario S2 in one embodiment of this application. Figure 9 This is a schematic diagram of the nonlinear relationship between the source-sink landscape and the rainfall time step under rainfall scenario S3 in one embodiment of this application. It is also a schematic diagram of the nonlinear relationship between the source-sink landscape and the rainfall time step under different rainfall scenarios, which shows the law of first rising and then falling in the source-sink conversion ratio during the rainfall process and its turning point. Figure 7 The nonlinear relationship between the source-sink ratio and the rainfall time step is described, and the inflection point is also shown. The fitting formula is as follows: . Figure 8 The nonlinear relationship between the source-sink ratio and the rainfall time step is shown under rainfall scenario S2, revealing a trend of the source-sink conversion ratio first increasing and then decreasing, with the inflection point marked. The fitted formula is: . Figure 9 This paper presents a schematic diagram illustrating the nonlinear relationship between the source-sink landscape and the rainfall time step under rainfall scenario S3, as well as the nonlinear relationship between the source-sink ratio and the rainfall time step. The fitting formula is as follows: The results of these schematic diagrams reflect the nonlinear relationship between source-sink landscape and rainfall time step under different rainfall scenarios. The existence of the nonlinear relationship indicates the dynamic change of the source-sink conversion ratio during the rainfall process, specifically manifested as a pattern of first increasing and then decreasing, and the corresponding inflection points are marked.

[0099] Figure 10 This diagram illustrates the identification of key three-dimensional landscape thresholds in one embodiment of this application, and shows the results of the key three-dimensional landscape threshold identification. For landscapes restored from source to sink, the value range of their key three-dimensional indicators can be obtained, thereby clarifying the three-dimensional indicator thresholds on different land use units. Specifically, plot A (S1) has an average leaf area index of 2.02 m² / m², an average three-dimensional green volume of 214.95 m³, and an average canopy coverage of 66.6%; plot B (S2) has an average leaf area index of 2.08 m² / m², an average three-dimensional green volume of 228.26 m³, and an average canopy coverage of 70.43%; and plot C (S3) has an average leaf area index of 2.10 m² / m², an average three-dimensional green volume of 236.74 m³, and an average canopy coverage of 74.66%. Figure 10 The results reflect the value range of key three-dimensional indicators during landscape restoration and annotate the values ​​of key three-dimensional indicators on different land use units. This helps to identify landscape features that play a key role in source-sink landscape transformation.

Claims

1. A method for analyzing surface runoff sources and sinks in urban three-dimensional landscapes, comprising: Acquire point cloud data, meteorological and hydrological data, and geospatial data, and preprocess the acquired data; Extract vegetation outline data and building outline data from preprocessed geospatial data; Based on the extracted vegetation outline data and building outline data, vegetation point cloud data and building point cloud data are extracted from the preprocessed point cloud data, respectively. Based on the extracted vegetation point cloud data and building point cloud data, calculate the three-dimensional indicators of vegetation and the three-dimensional indicators of buildings. Based on the preprocessed meteorological and hydrological data, rainfall-runoff simulation analysis was conducted using a hydrological model to obtain surface runoff depth data and calculate the source-sink area ratio and its inflection point. The values ​​of three-dimensional vegetation and building indices at the inflection point of surface runoff source-sink landscape obtained from simulation analysis are calculated and used as key three-dimensional indicators of surface runoff response. Rainfall-runoff simulation analysis using hydrological models includes: Construct a distributed hydrological model with a grid size of N. The distributed hydrological model includes a surface runoff sub-model, a saturated zone sub-model, an unsaturated zone model, and a river channel sub-model. Rainfall-runoff simulation analysis using hydrological models includes: The rainfall intensity formula is used to calculate the rainfall amount at different return periods. The rainfall intensity formula is shown below: ； Represents rainfall intensity in mm / min. Represents the design return period for rainfall. The duration of the designed rainfall (min) is represented by A1, C, n, and b, which are parameters determined according to the rainfall intensity formula for different regions. Based on the calculated rainfall amounts with different return periods, a rainfall process line corresponding to the return period is set as the rainfall input scenario, where the rainfall process line represents the rainfall intensity-time distribution.

2. The surface runoff source-sink analysis method for urban three-dimensional landscapes according to claim 1, characterized in that: Extract vegetation point cloud data from the preprocessed point cloud data, including: Use the preprocessed point cloud data as input data; Point cloud data that overlaps with the spatial distribution of vegetation outline range is obtained from the input data and used as the original vegetation point cloud data. Calculate the height, area, and volume of the original vegetation point cloud data to obtain the three-dimensional structure data of the vegetation point cloud. Based on the three-dimensional structure data of vegetation point cloud, leaf area index, canopy height, canopy coverage, canopy height heterogeneity, and green space parameters are calculated as three-dimensional vegetation indicators.

3. The surface runoff source-sink analysis method for urban three-dimensional landscapes according to claim 1, characterized in that: Extract building point cloud data from the preprocessed point cloud data, including: Use the preprocessed point cloud data as input data; Extract point cloud data that overlaps with the spatial distribution of the building outline from the input data, and use it as the original point cloud data of the building; Calculate the height, area, perimeter, and volume of the original building point cloud data to obtain the three-dimensional structural data of the building point cloud; Based on the 3D structural data of building point clouds, the roof height, slope, volume ratio, and crowding degree of the building point clouds are calculated as 3D indicators of the building.

4. The surface runoff source-sink analysis method for urban three-dimensional landscapes according to claim 1, characterized in that: The acquired meteorological and hydrological data are preprocessed, including: Different rainfall events are determined by the interval between rainfall events, and hourly rainfall data and hourly flow data are obtained. Hourly rainfall data represents the amount of rainfall per unit time, and hourly flow data represents the river flow per unit time.

5. The method for surface runoff source-sink analysis for urban three-dimensional landscapes according to claim 1, characterized in that: The acquired geospatial data is preprocessed, including: The acquired geospatial data is projected and transformed using the Albers equal-area conic projection to generate vector map data; Geographic coordinate registration is performed on the acquired geospatial data using known control points to generate registered image data; The generated vector map data and the registered impact data are correlated to form preprocessed geospatial data.

6. The method for surface runoff source-sink analysis for urban three-dimensional landscapes according to claim 1, characterized in that: Rainfall-runoff simulation analysis using hydrological models includes: Using the grid data of the surface runoff sub-model and the river sub-model as input, the spatial overlap algorithm is used to extract the spatial intersection grid of the surface runoff sub-model and the river sub-model as the water exchange connection unit. Based on the coupling relationship between surface runoff and river runoff processes, river water balance equations and surface water balance equations are established on the water exchange connection unit. The established river water balance equation and surface water balance equation are used as inputs to the equation set. The numerical calculation method of Saint-Venant equation set and the finite element discretization method are used to perform numerical calculation and mesh discretization on the equation set to obtain the coupled calculation results of the river hydrodynamic model and the surface hydrological model, which are used as the hydrological model.

7. The surface runoff source-sink analysis method for urban three-dimensional landscapes according to claim 6, characterized in that: Rainfall-runoff simulation analysis using hydrological models includes: Rainfall with different return periods was used as input data and input into the established hydrological model. Hourly rainfall and hourly flow data were used as calibration data to calibrate the hydrological model. The sensitivity indexes of the calibrated hydrological model are obtained using a global sensitivity algorithm. The Monte Carlo sampling method is used to obtain the value combination of sensitive indicators. The value combination is then input into a calibrated hydrological model for simulation calculation to obtain the model output results. Compare the statistical characteristic indicators of the model output results with those of the calibration data, select the sensitive indicators corresponding to the value combinations of statistical characteristic indicators that meet the threshold, and use the selected sensitive indicators as the optimal parameters of the hydrological model.

8. The method for surface runoff source-sink analysis for urban three-dimensional landscapes according to claim 7, characterized in that: Rainfall-runoff simulation analysis using hydrological models includes: The rainfall input scenarios corresponding to different return periods are used as inputs to import the hydrological model that determines the optimal parameters. For each rainfall input scenario, surface runoff depth data under different rainfall steps are calculated using a hydrological model with determined optimal parameters; Using the calculated surface runoff depth data under different rainfall steps, a surface runoff distribution map with the corresponding return period is generated; Using surface runoff distribution maps, green infrastructure is divided into source areas and sink areas, where source areas represent units that generate runoff and sink areas represent units that do not generate runoff. Calculate the area ratio of the source region and the sink region, and establish a polynomial model based on the rainfall duration and the area ratio of the source and sink regions; The threshold regression algorithm is applied to calculate the turning point in the dynamic change process of the source region based on the established polynomial model. The turning point represents the point where the slope changes abruptly before and after the proportion of the sink region to the source region changes. At the turning points in the dynamic changes of the source and sink regions, the corresponding three-dimensional indicators of vegetation and buildings are obtained as key three-dimensional indicators of surface runoff response.

9. The method for surface runoff source-sink analysis for urban three-dimensional landscapes according to claim 8, characterized in that: Key three-dimensional indices for calculating surface runoff response include: Surface runoff response indexes are calculated using simulated surface runoff depth data. These indexes include runoff volume and runoff length index. The correlation coefficients between surface runoff response indices and three-dimensional vegetation indices are calculated to obtain vegetation landscape correlation coefficients. Calculate the correlation coefficient between surface runoff response index and building three-dimensional index to obtain the building-landscape correlation coefficient; At the turning point in the dynamic change process of the source and sink regions, the vegetation landscape correlation coefficient and building landscape correlation coefficient corresponding to the turning point are obtained as three-dimensional landscape indices. The acquired three-dimensional landscape indexes will be used as key three-dimensional indicators of surface runoff response.

Images