Urban garden greening ecological benefit intelligent evaluation method based on digital twinning

Abstract

The invention discloses an urban landscaping ecological benefit intelligent evaluation method based on digital twinning, and particularly relates to the technical field of smart city and ecological information crossing, which comprises the following steps: collecting vegetation multispectral images, canopy three-dimensional point cloud, soil temperature and humidity and atmospheric microclimate parameters in real time, and combining a landscaping geometric model and a plant physiological parameter library to evaluate the ecological benefit of urban landscaping. The method comprises the following steps: constructing a high-fidelity and time-space synchronous urban landscaping digital twinborn body, dynamically extracting a multi-dimensional characteristic index, inputting an ecological process dynamic coupling model, outputting an ecological benefit comprehensive evaluation index and a time-space distribution map thereof, and supporting virtual scene deduction and multi-objective optimization by a system. Greening configuration and maintenance decision suggestions considering ecological benefits and cost benefits are generated, and dynamic display and closed-loop intelligent evaluation are realized through a visual interaction platform. According to the invention, real-time, accurate and systematic evaluation and intelligent management of urban landscaping ecological benefits are realized.

Description

Intelligent Assessment Method for Ecological Benefits of Urban Greening Based on Digital Twins Technical Field

[0001] This invention relates to the field of smart city and ecological information intersection technology, and more specifically, to a method for intelligent evaluation of the ecological benefits of urban landscaping based on digital twins. Background Technology

[0002] With the acceleration of urbanization and the increasing prominence of ecological and environmental problems, how to scientifically, systematically, and dynamically evaluate the ecological benefits of urban landscaping has become an important issue for sustainable urban development and refined management.

[0003] Existing technical solutions typically use remote sensing images to extract vegetation indices, and combine them with static ecological parameter databases and statistical models to estimate individual functions of urban green spaces, such as carbon sequestration and cooling effects. This includes using multispectral satellite data to calculate the normalized vegetation index, combining it with empirical biomass formulas to estimate regional carbon storage, analyzing the cooling effect of green spaces through surface temperature inversion, and using GIS spatial analysis tools to visualize the assessment results.

[0004] However, in practical use, it still has some shortcomings, such as lagging data updates, reliance on periodic remote sensing images, making it difficult to achieve real-time dynamic monitoring; relatively simple evaluation indicators, focusing on vegetation cover or single ecological functions, lacking multi-dimensional and systematic comprehensive evaluation; the model mechanism is out of touch with the actual situation, with most models based on static parameters, making it difficult to reflect the dynamic coupling effect of plant physiological processes and environmental factors; and it lacks the ability to interact with reality and make predictions, making it unable to support scientific decision-making for greening planning, renovation and maintenance management. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an intelligent assessment method for the ecological benefits of urban landscaping based on digital twins, which addresses the problems raised in the background art through the following solutions.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent assessment method for the ecological benefits of urban landscaping based on digital twins, comprising: S1: designating the urban landscaping to be assessed as the target area, deploying an IoT sensor network and satellite / aerial remote sensing data, and collecting dynamic environmental data in real time, including vegetation multispectral images, canopy 3D point clouds, soil temperature and humidity, and atmospheric microclimate parameters; combining a preset landscaping geometric model and plant physiological parameter library, and using a 3D modeling engine and physics engine to construct a high-fidelity digital twin of urban landscaping that is synchronized in real time with the physical landscaping scene and has spatiotemporal continuity; S2: based on real-time data stream driving, calculating and extracting multi-dimensional dynamic feature indicators reflecting the ecological benefits of landscaping in the digital twin, including real-time carbon sequestration calculated based on vegetation index and biomass model, and evaporation rate calculated based on canopy structure model and transpiration mechanism. The system includes: S3: Inputting multi-dimensional dynamic characteristic indicators into a preset ecological process dynamic coupling model, and outputting a comprehensive ecological benefit assessment index and its spatiotemporal distribution map; S4: Based on digital twins and the ecological process dynamic coupling model, performing virtual scenario simulations of planning or renovation schemes, predicting the ecological benefit evolution trajectory of different schemes in the future time period, and generating green space optimization configuration and maintenance management decision suggestions based on multi-objective optimization algorithms; S5: Dynamically rendering and displaying the comprehensive ecological benefit assessment index, spatiotemporal distribution map, virtual scenario simulation results and optimization decision suggestions through a visualization interactive platform, and supporting users to adjust model parameters or input management commands through the interactive interface to form an intelligent assessment system.

[0007] The technical effects and advantages of this invention are as follows: 1. This invention integrates IoT sensors, remote sensing data, and high-fidelity digital twins to extract multi-dimensional ecological indicators such as carbon sequestration, cooling, biodiversity, and rainwater retention in real time. Based on a dynamic coupling model of ecological processes, it outputs a comprehensive evaluation index and spatiotemporal map, solving the problems of single indicators, data lag, and static evaluation in traditional methods, and significantly improving the systematicness and timeliness of the evaluation; 2. This invention simulates the evolution of ecological benefits of different vegetation configurations, engineering modifications, and maintenance schemes in future periods through virtual scene simulation based on digital twins, and combines multi-view... The standard optimization algorithm generates optimization suggestions that take into account both ecological and cost-effectiveness, enabling urban greening planning and management to shift from experience-driven to data and model-driven approaches, thus improving the scientific and forward-looking nature of decision-making. 3. This invention establishes a closed-loop mechanism of "evaluation-decision-implementation-feedback-optimization" through a visual interactive platform that adjusts parameters in real time and obtains feedback. The model is dynamically updated every 3 months based on new data, which not only enhances the system's practicality and interactivity but also ensures that the evaluation model continues to improve with data accumulation and environmental changes, realizing the dynamic, refined, and intelligent management of urban landscaping. Attached Figure Description

[0008] Figure 1 is a schematic diagram of the overall structure of the present invention; Figure 2 is a schematic diagram of the construction of the digital twin of the present invention; Figure 3 is a schematic diagram of the extraction of multi-dimensional indicators of the present invention; Figure 4 is a schematic diagram of the dynamic coupling model of the ecological process of the present invention; Figure 5 is a schematic diagram of the virtual simulation and decision optimization of the present invention. Detailed Implementation

[0009] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0010] This implementation method elaborates on the intelligent assessment method of urban landscaping ecological benefits based on digital twins. This implementation method takes the core urban area of ​​a provincial capital city in the northern temperate monsoon climate as the target area for case verification. The technical solutions involved are adapted to different climate zones, city sizes and greening types.

[0011] As shown in Figures 1 and 2, the intelligent assessment method for the ecological benefits of urban landscaping based on digital twins includes S1: The urban landscaping to be assessed is designated as the target area. An IoT sensor network and satellite / aerial remote sensing data are deployed to collect dynamic environmental data in real time, including vegetation multispectral images, canopy 3D point clouds, soil temperature and humidity, and atmospheric microclimate parameters. Combined with a preset landscaping geometric model and plant physiological parameter library, a high-fidelity digital twin of urban landscaping with spatiotemporal continuity is constructed using a 3D modeling engine and a physics engine, which is synchronized in real time with the physical landscaping scene.

[0012] It should be further explained that GIS technology was used to finely divide the target area, with 10m×10m as the basic grid unit. Each unit was labeled with its green space type (park green space, roadside green space, residential area green space, unit-affiliated green space, protective green space), dominant vegetation species (specific family, genus, and species of trees, shrubs, and herbs), soil type (sandy soil, loamy soil, clay soil, and specific subtypes), topographic slope (0°-5°, 5°-15°, and above 15°), greening construction year, and basic information on the current maintenance responsibility entity. A grid-level geographic information index database was established. The data acquisition system adopted an "integrated air-space-ground" architecture. The specific data types and acquisition schemes are as follows: Airborne remote sensing data: multispectral data from the Gaofen-6 satellite was selected, with a spatial resolution of 2m, a spectral range of 450-950nm, and a revisit period of 5 days; UAV aerial photography data, equipped with a multispectral camera and lidar, with spectral bands including blue, green, red, near-infrared, and red-edge bands, and lidar point cloud density ≥50 points / m². 2 Data acquisition plan: Satellite data will be acquired every 5 days, drones will collect data on key areas once a month, and supplementary data will be collected within 48 hours after extreme weather events.

[0013] Ground-based IoT sensor data: Deploying three types of sensor nodes to form a sensor network combining grid coverage and focused monitoring. Vegetation monitoring nodes: every 200m 2 One monitoring station is deployed in the green space, equipped with a chlorophyll meter (measurement range 0-100 SPAD), a leaf temperature sensor (accuracy ±0.1℃), and a stem flow meter (measurement range 0-1000 mL / h) to monitor vegetation physiological activity. Three replicate monitoring stations are deployed in each soil type area at depths of 10cm, 30cm, and 50cm, equipped with temperature and humidity sensors (temperature accuracy ±0.2℃, humidity accuracy ±1%), conductivity sensors (accuracy ±5%), and soil moisture potential energy sensors. Atmospheric microclimate monitoring stations are deployed at key locations along the edge and within the green space, equipped with sensors for temperature (±0.1℃), humidity (±1%), wind speed (±0.1m / s), and photosynthetically active radiation (±5μmol / m²). 2 •s) and PM2.5 / PM10 sensors (accuracy ±1μg / m 3Data collection scheme: Vegetation and soil sensors collect data every 10 minutes, and atmospheric sensors collect data every 5 minutes. The data is transmitted in real time to edge computing nodes via a 5G-NB-IoT network.

[0014] Basic geometric and physiological parameter data: Basic data for the geometric model of the landscape greening were obtained through field surveys. Using a handheld laser rangefinder combined with high-definition photogrammetry, data on individual tree diameter at breast height (DBH) (accuracy ±0.1cm), tree height (±0.2m), crown width (±0.1m), branching point height (±0.1m), leaf number (per unit area), and bark thickness (±0.05cm) were collected to establish individual tree vegetation geometric profiles. The plant physiological parameter database adopted a "local measurement + literature supplementation + dynamic update" model, measuring the core parameters of 30 dominant vegetation species (sorted by quantity) within the target area, including light saturation point, maximum photosynthetic rate, and transpiration rate. The data, including the number of light compensation points, leaf stomatal conductance, and dark respiration rate, are supplemented with basic data from the "List of Garden Plants in Major Chinese Cities". The parameter values ​​are dynamically corrected based on quarterly monitoring data. Each quarter, leaf physiological samples of 30 dominant vegetation species in the target area are collected to determine the actual values ​​of the core parameters. The deviation rate between the measured values ​​and the original values ​​in the parameter database is calculated. If the deviation rate exceeds 5%, the parameter database is updated directly with the measured values. If the deviation rate is between 1% and 5%, the measured values ​​and historical data are integrated using the moving average method to optimize the parameter values. If the deviation rate is less than 1%, the original parameter values ​​are maintained to ensure dynamic matching between the parameter database and the actual physiological state of the vegetation.

[0015] A preprocessing pipeline of "outlier removal - spatiotemporal alignment - data augmentation" is constructed: Outlier handling adopts a combined method of "3σ criterion + trend consistency test" to remove outliers from the sensor data. For example, when soil moisture data exceeds the field water holding capacity ±3σ range for the soil type and is inconsistent with the trend of adjacent node data, it is identified as an outlier and is completed using time series-based LSTM interpolation; for cloud and shadow areas in remote sensing images, a linear hybrid decomposition method is used for restoration, preserving the spectral characteristics of vegetation pixels.

[0016] Using UTC time as the benchmark, a unified timestamp system is established to match the time scales of satellite data (5-day cycle), UAV data (monthly), and sensor data (5-10 minutes). Linear interpolation is used to interpolate low-frequency data to the 10-minute level. Spatially, a 10m×10m grid is used as the benchmark, and coordinate transformation (WGS84 to Gauss-Kruger projection) is used to achieve accurate alignment of remote sensing imagery, point cloud data, and GIS grid, with the alignment error controlled within 0.5m.

[0017] For vegetation multispectral data, a generative adversarial network (GAN) was used to generate simulated spectral data under different lighting conditions to improve the model's adaptability to lighting changes. For canopy point cloud data, a voxelization method was used to convert the three-dimensional point cloud into a voxel mesh (resolution 0.1m×0.1m×0.1m), and voxel density, height quantile, canopy voxel duty cycle, average voxel height, and standard deviation of voxel vertical stratification were extracted to enhance the three-dimensional structure representation ability.

[0018] Unity 3D was chosen as the 3D modeling engine, combined with NVIDIA... The PhysX physics engine constructs the core framework of the twin, with the following specific steps: Based on preprocessed point cloud data, a Poisson reconstruction algorithm is used to generate a 3D mesh model of the canopy. The model details are optimized by combining geometric parameters obtained from field surveys to achieve accurate geometric reproduction of individual plants. The photosynthetic rate, transpiration coefficient, light saturation point, light compensation point, stomatal conductance, and dark respiration rate from the plant physiological parameter library are associated with the geometric model. Numerical simulation modules for vegetation photosynthesis, transpiration, dark respiration, nutrient absorption, and water transport are constructed in the PhysX engine. A data interface adaptation layer is developed to drive the twin with a 10-minute real-time data stream, so that the twin's vegetation growth status (dynamic increments of plant height, canopy width, and number of leaves), soil moisture changes (temperature and humidity at different depths, electrical conductivity, and water potential energy), microclimate parameters (temperature, humidity, wind speed, photosynthetically active radiation, and PM2.5 / PM10 concentration), soil nutrient content (nitrogen, phosphorus, and potassium content), and vegetation physiological activity (chlorophyll content and transpiration rate) are updated synchronously with the physical scene.

[0019] A "two-factor calibration" mechanism is introduced to ensure the fidelity of the twin: Static calibration: After modeling is completed, the canopy volume, vegetation coverage, canopy projection area, leaf area index, and single-plant biomass of 100 sample points are measured in the field and compared with the twin simulation results. Genetic algorithms are used to optimize the modeling parameters so that the geometric error is ≤3%; Dynamic calibration: Every 7 days, UAV point cloud data and sensor data are used to calibrate the vegetation growth status (such as plant height and canopy width increment) and physiological status (such as transpiration rate) of the twin. An error feedback model is established to correct the simulation parameters of the physics engine in real time, ensuring the high fidelity of the twin's long-term operation.

[0020] The specific process of establishing the error feedback model includes: calculating the absolute deviation between the simulated twin value and the measured value. and relative deviation In the formula For the first Twin simulation values ​​of each monitoring indicator To correspond to the measured values, deviation thresholds were set for growth status indicators (plant height, canopy width) and physiological status indicators (transpiration rate). The growth index threshold was set at 3%, and the physiological index threshold was set at 5%; finally, a parameter correction function was constructed. In the formula These are the initial simulation parameters for the physics engine. The deviation correction factor is 0.8 for growth indicators and 1.2 for physiological indicators. When the relative deviation... When the threshold is exceeded, parameter correction is automatically triggered; when the threshold is below, the original parameters are maintained.

[0021] As shown in Figure 3, S2: Based on real-time data stream driving, multi-dimensional dynamic characteristic indicators reflecting the ecological benefits of garden greening are calculated and extracted in the digital twin, including real-time carbon sequestration based on vegetation index and biomass model, transpiration cooling effect based on canopy structure model and transpiration mechanism, biodiversity support index based on species distribution model and habitat quality assessment, and rainwater storage efficiency based on surface runoff simulation and soil infiltration model.

[0022] It should be further explained that the method of "vegetation index inversion + biomass estimation + photosynthetic mechanism coupling" is used to realize the real-time calculation of carbon sequestration: the normalized vegetation index (NDVI) and the enhanced vegetation index (EVI) are fused, and the two indices are calculated using UAV multispectral data. An index fusion model adapted to vegetation type is adopted: a fusion coefficient is constructed. Correlation equation with vegetation leaf area index (LAI) In the formula For the first Measured leaf area index of the vegetation. The maximum leaf area index of all vegetation in the target area is then calculated using the formula. Calculate the comprehensive vegetation index; among them, trees have a high LAI value. The value ranges from 0.55 to 0.65, emphasizing the sensitivity of NDVI to high vegetation cover; the LAI for shrubs is moderate. The value ranges from 0.4 to 0.5, balancing the advantages of both indices; herbal LAI values ​​are low. The value ranges from 0.3 to 0.4, focusing on the EVI's ability to resist interference under low vegetation cover, and solving the problem of saturation or distortion of a single index under different vegetation cover.

[0023] A linear regression model was established based on a comprehensive vegetation index. The model was then trained using biomass data from 50 different vegetation types measured in the field (using the harvest method), resulting in a formula for estimating biomass per unit area. In the formula, For the first Biomass per unit area of ​​vegetation (unit: kg / m²) 2 ); For the first Comprehensive vegetation index (dimensionless). For the first The regression slope coefficient of vegetation reflects the sensitivity of the comprehensive vegetation index to biomass (the value ranges from 12.5 to 18.2 for trees, from 5.8 to 9.3 for shrubs, and from 1.2 to 3.5 for herbs). For the first The regression intercept coefficient of vegetation was used to correct the baseline biomass (values ​​ranged from -0.8 to 0.2 for trees, -0.3 to 0.1 for shrubs, and 0.02 to 0.15 for herbs); the coefficient of determination was used during model training. To determine the goodness-of-fit criteria, we ensured that the biomass estimation error was controlled within 10%.

[0024] The Farquhar photosynthetic mechanism model is coupled in twins, and real-time microclimate data are input: photosynthetically active radiation, air temperature, air humidity, CO2 concentration, wind speed and soil surface temperature. The carbon sequestration rate per unit time and per unit biomass is calculated. Combined with the biomass estimation results, the real-time carbon sequestration amount at the 10-minute level is obtained, and the total daily carbon sequestration is summarized daily.

[0025] A three-layer extraction method combining "canopy structure analysis + transpiration mechanism simulation + microclimate response" was employed. Based on preprocessed point cloud voxel data, structural parameters were extracted: leaf area index (LAI), mean leaf tilt angle, canopy porosity, canopy height quantiles (25% / 50% / 75th percentiles), canopy volume density, and canopy projected coverage. LAI was calculated using the voxel density method with an accuracy of ±0.1m. 2 / m 2 A dual-source evapotranspiration model (separating vegetation transpiration from soil evaporation) was constructed in the twin model. Leaf area index (LAI), leaf temperature, soil moisture content, air temperature and humidity, wind speed, photosynthetically active radiation, and vegetation transpiration coefficient were input. Combined with transpiration coefficients from a plant physiological parameter database, the real-time transpiration rate per unit area of ​​vegetation was calculated. In the formula, The real-time transpiration rate of vegetation per unit area (unit: mm / h). The transpiration coefficient of the target vegetation is dimensionless and taken from the plant physiological parameter database. The value ranges from 0.6 to 0.9 for trees, 0.4 to 0.7 for shrubs, and 0.2 to 0.5 for herbs. The leaf area index (LAI) is the canopy leaf area index (taken from point cloud voxel extraction parameters). The saturated vapor pressure on the blade surface (unit: kPa, calculated from the blade temperature, formula: ...) , For the blade temperature; The actual water vapor pressure of the air (unit: kPa, calculated from air temperature and humidity data); Aerodynamic drag (unit: s / m) is negatively correlated with wind speed, and the formula is: , For wind speed, In order to monitor altitude, (for surface roughness). The value represents the stomatal resistance of the leaf (unit: s / m), which is negatively correlated with stomatal conductance and is taken from the plant physiological parameter database.

[0026] By using twin simulations to examine the microclimate differences with and without vegetation cover, a correlation model between transpiration rate and temperature drop was established. The specific formula is as follows: In the formula, The temperature drop of green space relative to bare, unvegetated land (unit: °C). >0 indicates a cooling effect); For the regression constant term, - The regression coefficients for each variable (obtained by fitting field monitoring data, with values ​​ranging from [value range to value range]) are as follows: -0.5 -0.2 2.8-4.5 : 0.6-1.1 : 1.2-2.0, 3.5-5.0; Real-time transpiration rate of vegetation per unit area (unit: mm / h); The leaf area index is the canopy area index. Wind speed (unit: m / s); Relative humidity (unit: %).

[0027] The model was calibrated using on-site measurements of temperature differences inside and outside the green space (20 comparison points were selected and monitored continuously for 7 days), with a coefficient of determination R. 2 A score of ≥0.88 is used as the goodness-of-fit criterion; based on this, an index of transpiration cooling effect is constructed. The calculation formula is: In the formula, The actual temperature drop after model calibration (unit: °C); The baseline temperature difference for bare land without vegetation in the target area (the average value within the monitoring period, unit: ℃); the larger the value of the index, the higher the cooling efficiency of the vegetation per unit transpiration rate, and the dynamic update is achieved at the 10-minute level.

[0028] A three-dimensional assessment system of "habitat quality-species association-ecological connectivity" is constructed: based on soil type, microclimate (temperature, humidity, wind speed, photosynthetically active radiation, CO2 concentration), vegetation cover, canopy leaf area index, soil moisture content, and topographic slope of twins, the habitat quality module of the InVEST model is used to calculate the habitat suitability score (0-10 points) for each grid. The specific calculation logic is as follows: determine the habitat type of the target area (tree forest, shrub forest, herbaceous green space), and assign a habitat suitability benchmark value to different habitat types. (8-10 points for arbor forests, 6-8 points for shrub forests, and 4-6 points for herbaceous green areas); Identify habitat stress factors and calculate the intensity of single-factor stress. In the formula The stress factor attenuation coefficient, Let be the distance from the grid to the stress source. Thresholds for the influence of stress factors; weighting the sensitivity of different habitat types to stress factors. Calculate the comprehensive stress index The final habitat suitability score formula is: In the formula This is a normalization constant with a value of 2.5; The value is a half-saturation constant, which is 0.5. A higher score indicates a better habitat quality.

[0029] A database of symbiotic relationships between 30 dominant vegetation species and 15 common associated animal species in the target area was established. The potential distribution probability of species under different vegetation configurations was analyzed using a species distribution model (MaxEnt). An ecological corridor network was constructed based on green space type and vegetation cover. The corridor connectivity index was calculated using a gravity model. The specific calculation logic was as follows: Ecological source areas within the target area were delineated: selected areas with vegetation cover ≥70% and an area ≥0.5 hm². 2 Arbor forests and shrub forests are designated as core ecological source areas, with each source area assigned a weight based on habitat quality. Based on the aforementioned habitat suitability score Normalized (values ​​range 0-1), ecological corridors are extracted: the minimum cost path between source areas is taken as the potential ecological corridor. The cost resistance coefficient is determined comprehensively based on green space type (resistance value 1 for arbor forest, 3 for shrub forest, 5 for herbaceous green space, and 20 for hard paving) and vegetation cover (the resistance coefficient increases by 1.2 times for every 10% decrease in cover). A gravity model is constructed to calculate the corridor connectivity index, using the following formula: In the formula, As the source With source The corridor connectivity index between corridors; the larger the value, the stronger the corridor connectivity. The cost distance of the corridor between source areas; The distance attenuation coefficient is set to 0.02; the overall connectivity index of the region is calculated. To assess the overall connectivity effectiveness of the ecological corridor network.

[0030] The weights of habitat quality (0.4), species distribution probability (0.3), and connectivity index (0.3) were determined using the analytic hierarchy process (AHP), and the weighted average was used to obtain the biodiversity support index (0-10 points), which was updated monthly.

[0031] A coupled approach of "surface runoff simulation + soil infiltration calculation + storage quantification" is adopted, and soil moisture potential energy parameters are introduced to improve accuracy: an improved SCS-CN model is constructed in a twin, and CN values ​​are determined by combining preprocessed soil type data (sandy soil, loam, clay). The specific calculation logic is as follows: based on the permeability classification of soil type (high permeability for sandy soil, medium permeability for loam, and low permeability for clay) and green space type (tree forest, shrub forest, herbaceous green space), the basic CN values ​​are determined with reference to the "SCS-CN Model Parameter Manual": tree forest (sandy soil 60-65, loam 70-75, clay 80-85), shrub forest (sandy soil 65-70, loam 75-80, clay 85-90), herbaceous green space (sandy soil 70-75, loam 0-85, clay 90-95); canopy interception is also introduced. , Leaf area index, the amount of rainfall during the period. At that time, the CN value is taken as 0.8 times the baseline value (without surface runoff). At that time, the CN value is calculated according to The scale is linearly adjusted to the base value.

[0032] Based on the real-time soil moisture content of the twins Classification of dry and wet conditions, drought CN value decreased by 10%, humidity Increase the value by 8% for moderate humidity (20%-40%), and maintain the baseline value; substitute the corrected CN value into the formula. , , To accurately simulate the flow of green land by measuring the potential retention capacity of the soil.

[0033] Real-time rainfall data (obtained by fusing data from atmospheric sensors and meteorological stations) is input to simulate surface runoff. Based on soil moisture potential energy sensor data, a dynamic relationship model between soil moisture content and permeability coefficient is established. The specific calculation logic is as follows: soil texture types are classified (sandy soil, loam, clay), and the saturated permeability coefficient of different soil textures is determined. Benchmark value, sandy soil =10-20mm / h, loam =2-5mm / h, clay =0.1-0.5 mm / h; introduce soil relative moisture content In the formula This represents the real-time soil volumetric water content. Soil wilting moisture content, To determine the soil saturation water content, a dynamic calculation formula for the permeability coefficient is constructed: In the formula, Real-time soil permeability coefficient (unit: mm / h); For soil texture correction factors, take 3.0-3.5 for sandy soil, 4.0-4.5 for loam, and 5.0-5.5 for clay; when hour, Take the minimum value of 0.05 mm / h (soil moisture deficit, extremely weak infiltration capacity); when hour, (Soil saturation, permeability reaches peak); combined with real-time monitoring of soil moisture content data by twins, the permeability coefficient is dynamically updated at the 10-minute level, supporting accurate simulation of green space flow and seepage processes.

[0034] This method replaces the traditional method of fixing the permeability coefficient, improving the accuracy of infiltration calculation. It uses twins to simulate the entire process of rainwater infiltration, soil water storage, and surface runoff, calculating the storage capacity (rainfall - runoff - evaporation), and combining this with the green space area to obtain the rainwater storage efficiency per unit area (unit: m²). 3 / 100m 2 The data is updated in real time after a rainfall event. The specific steps are as follows: determine the interception amount based on the canopy leaf area index (LAI). , For trees, the interception factor is 0.2 mm∙m. -2 Shrubs, take 0.15 mm∙m -2 Herbal extract 0.1 mm∙m -2 In the initial stage of rainfall, the amount of rainfall intercepted by the canopy is consumed first; after interception, the remaining rainfall... Combined with real-time soil permeability coefficient Calculate infiltration rate ( (This refers to the duration of rainfall), the infiltration water replenishes the soil's water storage, and the soil moisture content reaches saturation. At that time, excess water is converted into surface runoff; the rainwater retention capacity of green spaces In the formula This represents the total rainfall. Surface runoff, Evapotranspiration during the rainy season; Rainwater retention efficiency per unit area , Green space area, unit: m 2 The final result is output in mm.

[0035] As shown in Figure 4, S3: Input the multi-dimensional dynamic characteristic indicators into the preset ecological process dynamic coupling model, and output the comprehensive ecological benefit assessment index and its spatiotemporal distribution map.

[0036] It should be further explained that the input to the dynamic coupling model of ecological processes consists of four core indicators extracted by S2 (real-time carbon sequestration, transpiration cooling effect, biodiversity support index, and rainwater retention efficiency), as well as auxiliary parameters (vegetation configuration parameters, climate scenario parameters, and management intervention parameters). Among these, vegetation configuration parameters include the proportion of vegetation species, canopy closure, the proportion of tree-shrub-grass stratification, average canopy height, leaf area index, and seasonal vegetation phase characteristics; climate scenario parameters are based on the IPCCCMIP6 climate model, generating monthly average temperature and rainfall data for different emission scenarios (SSP1-2.6, SSP2-4.5, SSP5-8.5) for the next five years; management intervention parameters include irrigation frequency, irrigation volume, pruning cycle, pruning height, fertilizer application rate, fertilizer type, pest and disease control frequency, and soil improvement measures. Five typical management schemes are set up, with the following details: Ecological priority management scheme: irrigation frequency is 15 days / time, irrigation volume is 20 L / m³. 2 The pruning cycle is 60 days per cycle, with a pruning height retaining 70% of the canopy, and a fertilizer application rate of 50g / m². 2 (Mainly using well-rotted organic fertilizer), fertilization frequency is twice a year, pest and disease control adopts biological control (such as introducing natural enemies), soil loosening cycle is once every 90 days, and organic matter addition is 1 kg / m³. 2 Landscape-first management plan: Irrigation frequency is 7 days / time, irrigation water volume is 30L / m³. 2 The pruning cycle is once every 15 days, with a uniform pruning height of 50% of the canopy, and a fertilizer application rate of 80g / m². 2 (NPK compound fertilizer ratio 1:1:1), fertilization frequency is 4 times per year, pest and disease control adopts biological + chemical synergistic control, soil loosening cycle is 30 days / time, organic matter addition is 0.5kg / m³. 2 Low-cost operation and maintenance management solution: Irrigation frequency is 30 days / time, irrigation water volume is 10L / m³. 2 The pruning cycle is once every 120 days, with the pruning height retaining 80% of the canopy, and the fertilizer application rate is 20g / m². 2 (Replacing chemical fertilizers with straw return to the field), fertilization frequency is once a year, pest and disease control is mainly based on physical methods (such as insect-attracting lamps), soil loosening cycle is once every 180 days, no additional organic matter is added; water-saving and high-efficiency management plan: irrigation frequency is once every 20 days, irrigation water volume is 15L / m³ 2 (Drip irrigation method), pruning cycle is 45 days / time, pruning height retains 60% of the canopy, fertilizer application rate is 40g / m². 2(Slow-release fertilizer), fertilization frequency is twice a year, pest and disease control adopts ecological regulation (such as intercropping with insect-repelling plants), soil loosening cycle is once every 60 days, and the amount of organic matter added is 0.8 kg / m³. 2 Conventional balanced management plan: Irrigation frequency is 10 days / time, irrigation water volume is 25L / m³. 2 The pruning cycle is 30 days per cycle, with a pruning height retaining 65% of the canopy, and a fertilizer application rate of 60g / m². 2 (Organic fertilizer + compound fertilizer ratio 7:3), fertilization frequency is 3 times per year, low-toxicity chemical agents are used for pest and disease control, soil loosening is done every 45 days, and the amount of organic matter added is 0.6 kg / m³. 2 .

[0037] Z-score standardization is used to convert indicators of different dimensions into standardized data with a mean of 0 and a standard deviation of 1. Positive indicators are directly standardized, while negative indicators are standardized after taking their reciprocals.

[0038] The model adopts a two-layer architecture of "mechanism module + machine learning fusion module": The mechanism module constructs mechanistic sub-models for four core ecological processes, simulating the intrinsic correlation between indicators. Specifically, it includes a carbon-water coupling sub-model, based on the photosynthesis-transpiration coupling mechanism, establishing a correlation equation between carbon sequestration and transpiration cooling effect, and quantifying the trade-off between "increased carbon sequestration - increased transpiration water consumption"; and a water-biodiversity coupling sub-model, based on the habitat water demand mechanism, simulating the promoting effect of rainwater retention efficiency on the biodiversity support index, where retention efficiency exceeds a threshold (e.g., 1.5 m in loam areas). 3 / 100m 2 After that, the promoting effect slows down; the carbon-biodiversity coupling sub-model, based on the positive correlation between vegetation carbon storage and species diversity, establishes a linear correlation model, while considering the trade-off between high carbon storage and low biodiversity caused by large-scale planting of a single tree species.

[0039] The machine learning fusion module uses a long short-term memory network enhanced with an LSTM-Attention mechanism as the fusion model, inputting standardized indicator data and the output of the mechanism module. The attention mechanism is used to automatically identify the importance of different indicators in different spatiotemporal scenarios, and the LSTM module is used to capture the time-series dynamic features of the indicators. The model training uses three years of historical data of the target area, including indicator data under different seasons, climate events, and management measures. The ratio of training set, validation set, and test set is 7:2:1. The root mean square error (RMSE) is used as the loss function, and the model is iteratively trained through the Adam optimizer to ensure that the model prediction error is ≤5%.

[0040] The model outputs a comprehensive ecological benefit assessment index ranging from 0 to 100 points. A higher score indicates better ecological benefits. The calculation rule is as follows: the dynamic weights of each indicator are obtained through the LSTM-Attention model, and the weighted sum is then linearly stretched to 0-100 points. The rationality of the dynamic weights is verified by on-site expert scoring (five experts in the fields of landscaping and ecological assessment are invited to score the importance of benefits for 100 sample points). The correlation coefficient between the weights and the expert scores is ≥0.85.

[0041] The spatiotemporal distribution map generation is achieved jointly by ArcGIS and Unity 3D: Spatially, it uses 10m×10m grids as units, mapping the comprehensive evaluation index of each grid to a GIS map to generate a heat map, with red-yellow-green gradients representing increasing benefits; Temporally, it generates 10-minute dynamic maps (updated in real time), daily / weekly / monthly statistical maps, and predictive maps under different climate scenarios and management schemes. The maps accurately locate areas with weak ecological benefits and support interactive operations such as zooming, panning, and regional queries.

[0042] As shown in Figure 5, S4: Based on the dynamic coupling model of digital twin and ecological process, virtual scenario simulation is performed on the planning or transformation scheme to predict the ecological benefit evolution trajectory of different schemes in the future time period. Based on the multi-objective optimization algorithm, decision suggestions for green space optimization configuration and maintenance management are generated.

[0043] It should be further explained that the input interface for constructing schemes based on digital twins supports users to input three types of scheme parameters: vegetation configuration schemes include the types of vegetation to be added / replaced, variety selection, seedling specifications (plant height, ground diameter, crown width), planting quantity, planting density, planting location, vertical stratification ratio of trees, shrubs and grasses, and seasonal vegetation matching (perennial / deciduous, flowering / foliage plant ratio), providing a list of suitable vegetation for the target area and descriptions of its growth characteristics; engineering transformation schemes include terrain adjustment (slope, elevation), soil improvement (adding organic matter, improving sand-clay ratio), and irrigation system upgrades (drip irrigation / sprinkler irrigation switching); maintenance and management schemes include irrigation frequency, pruning cycle, and fertilizer application. The system has 10 built-in typical scheme templates, and users can generate custom schemes by modifying parameters based on the templates.

[0044] The scheme parameters are input into the digital twin and the dynamic coupling model of ecological processes to conduct virtual scenario simulations for the next 1-5 years, with a prediction time step of one month. During the simulation, the twin simulates the vegetation growth process, soil state evolution, and microclimate response after the implementation of the scheme in real time. The ecological process model outputs a comprehensive ecological benefit assessment index for each month based on dynamic changes, forming an evolution trajectory curve. Monte Carlo simulation is used to conduct uncertainty analysis to improve the reliability of the prediction. The influence of climate fluctuations and vegetation growth variations is considered to generate an evolution trajectory with a 95% confidence interval.

[0045] The optimization objectives were determined to be "maximizing ecological benefits" and "minimizing costs." Costs included initial investment costs (vegetation procurement and engineering modification costs), for which a cost database was established based on local building material and seedling market prices; and operation and maintenance costs (irrigation, pruning, and fertilization costs), which were calculated by combining scheme parameters with local water, electricity, and fertilizer prices. The non-dominated sorting genetic algorithm NSGA-Ⅲ was used for multi-objective optimization, with the algorithm parameters set as follows: population size 100, number of iterations 50, crossover probability 0.8, and mutation probability 0.05.

[0046] After optimization, a Pareto optimal solution set (10-15 preferred solutions) is generated. Each solution is then comprehensively evaluated: the ecological benefit improvement (compared to the current state) is calculated relative to the cost input, yielding the benefit-cost ratio (BCR). The specific calculation logic is as follows: the core ecological benefit indicators of the green space after the implementation of the solution are quantified, and each indicator is converted into a uniform monetary value using the ecosystem service value assessment method to obtain the total ecological benefit value. In the formula For the first The physical quantity of ecological benefits. This represents the unit value coefficient of the benefit; the percentage increase in benefit compared to the current situation. In the formula The total ecological benefit value of existing green spaces; the total lifecycle investment cost of the statistical plan implementation. ,in For the cost of seedling procurement and planting, For routine maintenance costs such as irrigation, pruning, and fertilization, Costs related to pest and disease control and soil improvement. Costs of vegetation renewal and replanting; calculation of benefit-cost ratio ,when This indicates that the ecological benefits of the plan outweigh the cost input, and the plan is economically feasible. The higher the value, the better the overall benefits of the solution.

[0047] Combining expert scores (scored from three dimensions: ecological stability, landscape effect, and operability), the TOPSIS method was used to rank the Pareto solutions and select the optimal solution. The final decision recommendations include an optimal vegetation configuration list (species, quantity, and planting location); specific engineering transformation measures and implementation sequence; maintenance and management details (irrigation time, pruning intensity, and fertilization type); and a benefit prediction report (ecological benefit evolution trend, cost details, and BCR value over the next three years).

[0048] S5: The comprehensive ecological benefit assessment index, spatiotemporal distribution map, virtual scenario simulation results, and optimization decision-making suggestions are dynamically rendered and displayed through a visual interactive platform. Users can also adjust model parameters or input management commands through the interactive interface to form an intelligent assessment system.

[0049] It should be further explained that a multi-terminal collaborative architecture of "Web + Mobile + Immersive VR" is adopted. The Web and Mobile terminals are developed based on the React framework, and the VR terminal is developed based on Unreal Engine 5, enabling multi-scenario display of evaluation results: The data dashboard module displays the real-time ecological benefit comprehensive evaluation index, the values ​​and trend curves of four core indicators, and supports data export (Excel / PDF format); the spatiotemporal mapping module loads a 10m×10m precision spatiotemporal distribution heat map, supports dragging the timeline to view the changes in the map over different periods, and clicking on the grid to view detailed indicator data; the VR virtual simulation module allows users to enter the digital twin from a first-person perspective and intuitively view the implementation effects of different schemes (such as the landscape effect after vegetation growth and microclimate changes), and the Web and Mobile terminals support the display and comparison of 3D models of scheme effects; the decision suggestion module displays optimization decision suggestions in the form of text, images and tables, and associates the virtual scenes and benefit prediction curves corresponding to the schemes.

[0050] Develop an interactive workflow of "parameter adjustment - real-time simulation - effect feedback": Users modify model parameters or input management commands through the interactive interface. The system verifies the rationality of the parameters in real time and provides parameter adjustment suggestions. After the parameters are submitted, the digital twin and the ecological process model complete a single simulation within 10 seconds, outputting the adjusted ecological benefit assessment results and scenario display. The system generates a benefit comparison report before and after parameter adjustment, clarifying the correlation between "adjustment and benefit change".

[0051] Constructing a long-term closed-loop mechanism: The data acquisition closed loop records the management measures implemented by users as management parameters, synchronizes them to the twin, and verifies the effectiveness of the scheme implementation by combining them with subsequently collected sensor and remote sensing data; the model optimization closed loop updates the dynamic coupling model of the ecological process every 3 months based on newly collected data and the effectiveness of scheme implementation, using incremental learning methods to improve the accuracy of assessment and prediction; the decision iteration closed loop optimizes typical scheme templates based on annual assessment results and user feedback, forming a long-term closed-loop intelligent system of "assessment-decision-implementation-feedback-optimization".

[0052] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other. Finally: The above description is only a preferred embodiment of this invention and is not intended to limit this invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this invention should be included within the protection scope of this invention.

Claims

1. A method for intelligent evaluation of the ecological benefits of urban landscaping based on digital twins, characterized in that: include: S1: The urban green space to be evaluated is designated as the target area. An IoT sensor network and satellite / aerial remote sensing data are deployed to collect dynamic environmental data in real time, including vegetation multispectral images, canopy 3D point clouds, soil temperature and humidity, and atmospheric microclimate parameters. Combined with a pre-defined geometric model of the green space and a plant physiological parameter library, a high-fidelity digital twin of the urban green space is constructed using a 3D modeling engine and a physics engine, ensuring real-time synchronization with the physical green space scene and spatiotemporal continuity. S2: Based on real-time data streams, multi-dimensional dynamic characteristic indicators reflecting the ecological benefits of the green space are calculated and extracted from the digital twin, including real-time fixed data calculated based on vegetation indices and biomass models. Carbon content, transpiration cooling effect calculated based on canopy structure model and transpiration mechanism, biodiversity support index based on species distribution model and habitat quality assessment, and rainwater storage efficiency calculated based on surface runoff simulation and soil infiltration model; S3: Input multi-dimensional dynamic characteristic indicators into a preset ecological process dynamic coupling model, and output a comprehensive ecological benefit assessment index and its spatiotemporal distribution map; S4: Based on digital twin and ecological process dynamic coupling model, perform virtual scenario simulation of planning or transformation schemes, predict the ecological benefit evolution trajectory of different schemes in the future time period, and generate green space optimization configuration and maintenance management decision suggestions based on multi-objective optimization algorithm; S5: The comprehensive ecological benefit assessment index, spatiotemporal distribution map, virtual scenario simulation results, and optimization decision-making suggestions are dynamically rendered and displayed through a visual interactive platform. Users can also adjust model parameters or input management commands through the interactive interface to form an intelligent assessment system.

2. The intelligent assessment method for the ecological benefits of urban landscaping based on digital twins as described in claim 1, characterized in that: The high-fidelity digital twin of the landscape greening system includes a gridded geographic information index library, using 10m×10m as the basic unit, labeling green space type, dominant vegetation species, soil type, terrain slope, construction year, and maintenance responsibility information; through an integrated air-space-ground data acquisition architecture, it integrates satellite multispectral data, UAV lidar point cloud data, and ground IoT sensor data to establish a preprocessing pipeline for outlier removal, spatiotemporal alignment, and data augmentation; based on Unity 3D and NVIDIA PhysX engine, it constructs the core framework of the twin to achieve dynamic coupling of vegetation geometric models and physiological parameters, and introduces a two-factor calibration mechanism, which includes a static calibration model and a dynamic error feedback model.

3. The intelligent assessment method for the ecological benefits of urban landscaping based on digital twins as described in claim 1, characterized in that: The calculation of the multi-dimensional dynamic characteristic indicators includes: using a vegetation index fusion model, combined with a leaf area index correlation equation adapted to vegetation type, to calculate a comprehensive vegetation index, and based on a biomass regression model and a Farquhar photosynthetic mechanism model; establishing an transpiration cooling effect index based on canopy structure analysis and a dual-source evapotranspiration model, combined with a microclimate response model, to dynamically quantify the cooling efficiency of vegetation; constructing a three-dimensional assessment system of habitat quality, species association, and ecological connectivity, combined with the InVEST model, the MaxEnt species distribution model, and a gravity model to calculate the biodiversity support index; and using an improved SCS-CN model and a dynamic infiltration model of soil moisture potential energy to simulate the rainwater storage process and evaluate the rainwater storage efficiency per unit area in real time.

4. The intelligent assessment method for the ecological benefits of urban landscaping based on digital twins as described in claim 1, characterized in that: The extraction and calculation of the transpiration cooling effect includes using a dual-source evapotranspiration model to separate vegetation transpiration from soil evaporation, and combining canopy leaf area index, leaf temperature, soil moisture content, air temperature and humidity, wind speed, photosynthetically active radiation and transpiration coefficient and stomatal resistance in the plant physiological parameter database to calculate the real-time transpiration rate of vegetation per unit area. By establishing a correlation model between transpiration rate and cooling amplitude, the microclimate differences with and without vegetation cover are simulated, and the model is calibrated using temperature difference data inside and outside green spaces monitored in the field. Based on the calibrated model, an index of transpiration cooling effect is constructed to reflect the cooling efficiency of vegetation per unit transpiration rate.

5. The intelligent assessment method for the ecological benefits of urban landscaping based on digital twins according to claim 1, characterized in that: The dynamic coupling model of ecological processes adopts a two-layer architecture of mechanism module and machine learning fusion module. The mechanism module includes carbon-water coupling sub-model, water-biodiversity coupling sub-model and carbon-biodiversity coupling sub-model, which are used to simulate the intrinsic correlation and trade-off between different ecological processes. The machine learning fusion module uses a long short-term memory network enhanced by the LSTM-Attention mechanism to dynamically identify the importance of each indicator in different spatiotemporal scenarios, and trains the model based on historical data to output a comprehensive ecological benefit evaluation index of 0-100 points; it also combines ArcGIS and Unity 3D to generate spatiotemporal distribution maps.

6. The intelligent assessment method for the ecological benefits of urban landscaping based on digital twins according to claim 1, characterized in that: The virtual scenario simulation includes supporting users to customize vegetation configuration, engineering modification, and maintenance management parameters through the scheme input interface, and the system has built-in multiple typical scheme templates; based on the dynamic coupling model of digital twins and ecological processes, it performs simulations over a time step of 1-5 months in the future, and combines Monte Carlo simulation to generate an ecological benefit evolution trajectory with a 95% confidence interval; it uses the non-dominated sorting genetic algorithm NSGA-III for multi-objective optimization, with the goal of maximizing ecological benefits and minimizing costs, and outputs a Pareto optimal solution set; it calculates the benefit-cost ratio through the ecosystem service value assessment method, and combines expert scoring and the TOPSIS method to select the optimal scheme, outputting decision recommendations including a vegetation configuration list, engineering modification measures, maintenance management details, and a benefit prediction report.

7. The intelligent assessment method for the ecological benefits of urban landscaping based on digital twins as described in claim 1, characterized in that: The visualization and interaction platform adopts a multi-terminal collaborative architecture of Web, mobile and immersive VR terminals, and supports real-time ecological benefit data dashboards, spatiotemporal distribution heat maps, 3D display of virtual scene simulations and text and graphic output of decision-making suggestions.

8. The intelligent assessment method for the ecological benefits of urban landscaping based on digital twins according to claim 1, characterized in that: The intelligent evaluation system includes an interactive mechanism of "parameter adjustment - real-time simulation - effect feedback", which allows users to adjust model parameters or input management commands in real time through the interactive interface. After the system completes a single simulation, it outputs the updated ecological benefit evaluation results. Generate a benefit comparison report before and after parameter adjustment. The benefit comparison report is used to quantify the impact of different parameters or management measures on ecological benefits and clarify the correlation between adjustment and benefit changes.

Images