Precipitation forecasting method and system based on spatial non-uniform fractal urban canopy model

Abstract

The application discloses a precipitation prediction method and system based on a spatial non-uniform fractal urban canopy model, and the precipitation prediction method based on the spatial non-uniform fractal urban canopy model comprises the following steps: collecting meteorological data of a target urban area; inputting the collected meteorological data into a mesoscale numerical weather prediction model fused with the spatial non-uniform fractal urban canopy model, and outputting a precipitation prediction result. The spatial non-uniform fractal urban canopy model is a construction of urban land surface parameters with a spatial scale scaling law characteristic; the fractal geometry theory is used to describe the spatial scale scaling law of the urban building distribution, and the theory is introduced into numerical weather prediction, so that the numerical weather prediction has the ability to solve the non-uniformity of the city on a fine scale on the meteorological mechanism across scales, especially the local atmospheric movement process that may cause strong convection and precipitation, and thus the simulation precision and the prediction ability of strong precipitation events in the city and the surrounding area are effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of urban low-altitude numerical weather prediction technology, specifically involving a precipitation forecasting method and system based on a spatially non-uniform fractal urban canopy model. Background Technology

[0002] The high-density building clusters resulting from current urbanization have created highly complex and heterogeneous surface structures. These structures significantly influence the urban boundary layer thermodynamic mechanisms by altering local energy balance and turbulent exchange processes, thereby exacerbating extreme weather events such as heavy precipitation. However, current mainstream mesoscale numerical models have fundamental limitations in their urban parameterization schemes: on the one hand, these models simplify the urban underlying surface into a uniformly parameterized ideal plane, using fixed roughness, albedo, and other values ​​to characterize surface properties. This results in an inability to depict the gradient of real building distribution in simulations, leading to systematic shifts in precipitation areas and lags in the timing of convection triggering. On the other hand, the inherent fractal characteristics of urban spatial structures, including but not limited to the self-similarity of building clusters and the anisotropy of street networks, should cause surface physical parameters to dynamically change with the grid scale according to spatial scaling laws. However, in current weather forecasting models, reproducing urban severe convective precipitation processes often requires varying the grid resolution (…). For example, unstructured grids or multi-resolution nested grid techniques can be used to downscale numerical simulations of atmospheric motion from meso-large to meso-small scales. However, existing nested grid techniques force the same parameter set to be used at all levels (e.g., 32km / 8km / 2km grids), resulting in cross-scale error accumulation problems such as coarse grids over-smoothing urban heat islands and fine grids failing to resolve micro-scale eddies. Traditional schemes ignore the regulatory role of real urban spatial heterogeneity on meteorological elements and fail to establish a dynamic response mechanism between parameters and grid resolution. This directly leads to the model's inability to accurately capture the development path of local severe convection in cities, manifesting as systematic biases such as shifts in simulated precipitation areas and underestimation of intensity, severely restricting the accuracy of heavy precipitation forecasts. The insufficient accuracy of urban precipitation simulation essentially reveals the significant limitations of existing models in characterizing the urban street-valley scale thermodynamic environment. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a precipitation forecasting method and system based on a spatially heterogeneous fractal urban canopy model, which effectively improves the simulation accuracy and forecasting capability of heavy precipitation events in cities and surrounding areas.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0005] Firstly, a precipitation forecasting method based on a spatially heterogeneous fractal urban canopy model is provided, comprising: collecting meteorological data of the target urban area; inputting the collected meteorological data into a mesoscale numerical weather prediction model that integrates the spatially heterogeneous fractal urban canopy model, and outputting precipitation forecast results.

[0006] Furthermore, the method for constructing a spatially heterogeneous fractal urban canopy model includes: acquiring remote sensing images or GIS building boundary data of the target urban area and performing gridding processing to obtain a surface grid, while dividing the surface grid into urban built-up areas and natural surfaces; acquiring key land surface parameters of the urban built-up area at the observation scale and obtaining the spatial scaling law characteristics of the key land surface parameters; calculating the fractal dimension of the target urban area based on fractal geometry methods; and establishing a spatially heterogeneous fractal urban canopy model based on the fractal dimension and the spatial scaling law characteristics of the key land surface parameters.

[0007] Furthermore, the surface grid is divided into urban built-up areas and natural surfaces, including: dividing the urban built-up areas into building rooftop areas and street areas; performing numerical statistical analysis on the spatial distribution characteristics of building rooftop areas and street areas based on patch recognition technology to obtain statistical analysis results; binarizing the statistical analysis results and marking the building rooftop areas and street areas respectively; wherein, the street areas include: roads and square land; the natural surfaces include: natural features within the urban built-up areas, including vegetation-covered areas, water bodies, and non-buildable mountainous areas.

[0008] Furthermore, the urban built-up area is composed of multi-level self-similar building clusters, including: "small areas" formed by several buildings and streets between buildings, i.e., primary clusters; "communities" formed by multiple "small areas" and streets between "small areas", i.e. secondary clusters; urban areas formed by multiple "communities" and streets or gaps between "communities", i.e. tertiary clusters; and multiple secondary clusters, tertiary clusters, and roads between clusters constitute the urban built-up area.

[0009] Furthermore, based on fractal geometry methods, the fractal dimension of the target urban area is calculated, specifically as follows:

[0010] The fractal dimension D of the urban built-up area is defined as:

[0011] ;

[0012] in, This represents the ratio of the feature lengths of the tertiary cluster block to the secondary cluster block, where N represents the ratio of the measurements of the building cluster blocks at two different spatial scales.

[0013] ;

[0014] ;

[0015] ;

[0016] L represents the characteristic length of the tertiary cluster, W represents the average width of the street and natural surface, p represents the average measure of the tertiary cluster, the number of secondary clusters in the tertiary cluster is n×n, and k represents the number of natural surfaces in the tertiary cluster.

[0017] Furthermore, the value of any key land surface parameter in a given surface grid is obtained by weighting the key land surface parameters of both the natural surface and the urban built-up area contained in that surface grid, as well as their respective area proportions within that surface grid, using weighting coefficients. That is:

[0018] ;

[0019] in, This indicates that the grid resolution is the observation scale. The average key land surface parameters within the surface grid are times higher. The value; This indicates the proportion of urban built-up area within the surface grid. This represents the numerical value of the key land surface parameter of the natural surface within the grid. This indicates that the grid resolution on the Earth's surface is the observation scale. When the value is multiplied, it represents the value of the same key land surface parameter in the urban built-up area at this characteristic spatial scale.

[0020] Furthermore, the fusion of the spatially heterogeneous fractal urban canopy model and the mesoscale numerical weather prediction model specifically involves: embedding the spatially heterogeneous fractal urban canopy model into the mesoscale numerical weather prediction model, and achieving cross-scale optimization simulation of the nested grid system through bidirectional coupling of the water and heat flux transport process between the land and atmosphere. This includes: embedding the spatially heterogeneous fractal urban canopy model into the urban submodule of the land surface process model of the mesoscale numerical weather prediction model, replacing the existing underlying surface parameter scheme; dynamically calculating urban morphological parameters based on grid resolution, including the planar index and the windward area index; and receiving the near-surface temperature, humidity, and wind speed fields output by the mesoscale numerical weather prediction model at each time step, generating sensible heat, latent heat, and momentum turbulence fluxes, and feeding them back to the atmospheric module.

[0021] Furthermore, the mesoscale numerical weather prediction model includes the WRF model and the CMAQ model.

[0022] Furthermore, the accuracy of precipitation forecasts is quantitatively assessed by calculating relative errors and correlation coefficients:

[0023] ;

[0024] ;

[0025] Where BIAS represents the relative error, and r represents the correlation coefficient. and These represent forecasted precipitation and observed precipitation, respectively. This represents the average value of the forecast precipitation. This represents the average observed precipitation, and n represents the number of data points.

[0026] Secondly, this invention provides a precipitation forecasting system based on a spatially heterogeneous fractal urban canopy model, comprising: a data acquisition module for collecting meteorological data of a target urban area; and a precipitation forecasting module for inputting the collected meteorological data into a mesoscale numerical weather prediction model that incorporates the spatially heterogeneous fractal urban canopy model, and outputting precipitation forecast results. Compared with existing technologies, the beneficial effects achieved by this invention are:

[0027] (1) This invention improves the simulation accuracy and forecasting capability of heavy precipitation events in cities and surrounding areas by inputting collected meteorological data into a mesoscale numerical weather prediction model that integrates a spatially heterogeneous fractal urban canopy model.

[0028] (2) This invention not only considers the regulatory effect of urban spatial heterogeneity on meteorological elements, but also establishes a dynamic response mechanism for parameters and grid resolution, accurately captures the development path of local strong convection in the city, and reveals the significant limitations of existing models in characterizing the urban street valley scale thermodynamic environment. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the main process of a precipitation forecasting method based on a spatially heterogeneous fractal urban canopy model provided in an embodiment of the present invention;

[0030] Figure 2 The flowchart shows the application of the spatially non-uniform fractal urban canopy model provided by this invention to the numerical weather prediction model (taking WRF as an example).

[0031] Figure 3 This is geographic information system (GIS) data for the building surface of a certain city;

[0032] Figure 4 This is a schematic diagram of the spatial self-similarity of cities without feature scale proposed in this invention, derived from urban GIS data (a) at 10km resolution; (b) at 1km resolution; (c) at 0.1km resolution;

[0033] Figure 5 This is a map of urban building distribution patterns that conform to spatial self-similar structural characteristics and have spatial non-uniformity, as proposed in this invention.

[0034] Figure 6 This is a binarized image of the spatial characteristics of the actual building distribution in a city, obtained from GIS data.

[0035] Figure 7 It is the spatiotemporal variation of key land surface parameters (taking average albedo as an example) within the urban surface grid (model of this invention vs. traditional model).

[0036] Figure 8 Comparative analysis of the results and observed values ​​of the experimental group (the model after embedding the spatially non-uniform fractal urban canopy model) and the control group (the original model). Detailed Implementation

[0037] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0038] Example 1

[0039] like Figures 1-8 As shown, a precipitation forecasting method based on a spatially heterogeneous fractal urban canopy model includes: collecting meteorological data of the target urban area; inputting the collected meteorological data into a mesoscale numerical weather prediction model that integrates the spatially heterogeneous fractal urban canopy model, and outputting precipitation forecast results. The method for constructing the spatially heterogeneous fractal urban canopy model includes: acquiring remote sensing images or GIS building boundary data of the target urban area and performing gridding processing to obtain a surface grid, while dividing the surface grid into urban built-up areas and natural surfaces; acquiring key land surface parameters of the urban built-up area at the observation scale and obtaining the spatial scaling law characteristics of the key land surface parameters; calculating the fractal dimension of the target urban area based on fractal geometry methods; and establishing a spatially heterogeneous fractal urban canopy model based on the fractal dimension and the spatial scaling law characteristics of the key land surface parameters.

[0040] I. Standardization of Urban Architectural Spatial Structure and Extraction of Fractal Features.

[0041] Step S1: Acquire remote sensing imagery or GIS building boundary data for the target urban area. Perform format conversion and vector-to-rasterization on the raw building boundary data to generate a standardized grid layer.

[0042] The surface grid is divided into urban built-up areas and natural surfaces, including: dividing the urban built-up areas into building rooftop areas and street areas; performing numerical statistical analysis on the spatial layout characteristics of building rooftop areas and street areas based on patch recognition technology to obtain statistical analysis results; performing binarization processing on the statistical analysis results and marking the building rooftop areas and street areas respectively; wherein, the street areas include: roads and square land; the natural surface includes: natural features within the urban built-up areas, including vegetation-covered areas, water bodies and non-buildable mountainous areas.

[0043] Construct an urban underlying surface with spatially heterogeneous structural characteristics.

[0044] First, remote sensing imagery and GIS building vector data are used to obtain information on the spatial distribution of buildings and street valleys on the urban underlying surface, and then a spatial geometric model of urban buildings and streets is reconstructed.

[0045] In urban surface modeling, most physical and mathematical characteristic parameters (such as surface roughness, heat capacity, albedo, etc.) in urban grid cells are usually processed by weighted averaging of built areas and natural areas (including vegetation, water bodies, etc.) according to their area ratio. Although this method is representative to some extent, it has certain limitations in depicting complex urban spatial structures.

[0046] To improve the accuracy of urban spatial pattern description, this invention proposes to clearly divide the urban area into urban built-up area and natural surface. The urban built-up area is composed of high-density artificial buildings and is further subdivided into building roof coverage area and street area. The street area includes roads and squares. The natural surface includes natural features within the urban built-up area, including vegetation cover area, water body and non-buildable mountain area.

[0047] Obtaining the spatial structural characteristics of a building area includes: for the urban built-up area, it can be distinguished into two categories based on geographic information system (GIS) data: those occupied by building rooftops and those occupied by streets. The original building boundary data within the GIS is then processed into standard grids to form basic raster-structured patch data, such as... Figure 3 As shown, the area covered by building rooftops is marked as black pixels, and the street area is marked as white pixels. The entire area can be viewed as several ordered clusters with random breaks, and within any one cluster, there are similar structures at the next scale. This structure, exhibiting self-similarity and an ordered-disorder spatial distribution, possesses fractal geometric characteristics.

[0048] Statistical analysis was conducted on the layout characteristics of building-street areas across different spatial scales to identify structural forms in which building areas are generally separated by blank areas.

[0049] Modeling is performed after incorporating topographic and water system disturbance features into the aforementioned spatial structure. Specifically, this includes obtaining parameters related to urban spatial heterogeneity, such as street width-to-height ratio (HWR), average building canopy height (H), main street orientation, and spatial clustering coefficient.

[0050] Based on information about mountains, water systems, or nature reserves within the urban area, the building distribution boundaries are modified to identify the asymmetry and irregularity of the building-non-building area boundaries.

[0051] The original building boundary data is standardized and gridded to construct a regular grid distribution structure. A certain proportion of asymmetric broken units are introduced into the regular grid, such as water bodies, urban green spaces, mountains and other areas that are not suitable for construction, in order to simulate the natural disturbance or interference characteristics of undeveloped areas in urbanized areas, thereby enhancing the model's ability to express the heterogeneous structure of urban space.

[0052] Step S2: Extract the structural features of the building layers at multiple spatial scales and identify that they exhibit spatial self-similarity at multiple scales, conforming to scale-free fractal structures.

[0053] Based on the above analysis results, it is determined that urban built-up areas have spatial self-similarity without characteristic scales at multiple scales, and this is described by fractal geometry, with the corresponding quantitative index being its fractal dimension D.

[0054] The urban built-up area is composed of multi-level self-similar building clusters.

[0055] Urban built-up areas are composed of clusters with certain self-similarity characteristics at different spatiotemporal scales. That is, at a smaller scale, a "small area" is formed by a number of buildings and streets (first-level streets), which is a primary cluster; a number of buildings and first-level streets form the smallest unit (small area).

[0056] A larger "community," or secondary cluster, is formed by multiple such "small areas" and wider streets (secondary streets, wider than the primary streets between buildings within the aforementioned "small areas"). Multiple primary clusters and secondary streets form community units, with the width of the secondary streets greater than that of the primary streets. Further, multiple such "communities" and wider streets (tertiary streets) or gaps between them form a larger urban area, or tertiary cluster. Multiple secondary clusters and tertiary streets or gaps constitute the urban built-up area, with the width of the tertiary streets greater than that of the secondary streets. The cluster-like urban canopy model with self-similar spatial characteristics proposed in this invention is closer to reality.

[0057] Step S3: Calculate the fractal dimension D of the region based on the box dimension equal fractal geometry method.

[0058] Based on fractal geometry methods, the fractal dimension of the target urban area is calculated as follows: Assuming that the characteristic lengths of building clusters, the street widths within them, and the number of building clusters contained therein (hereinafter referred to as measures) can be directly obtained through observation methods at a certain spatial scale, this is called the observable scale. Simultaneously, for another secondary building cluster region with geometrically self-similar characteristics, the fractal dimension D of the urban built-up area is defined as:

[0059] (1)

[0060] in, This represents the ratio of the feature lengths of the tertiary cluster block to those of the secondary cluster block, where N represents the ratio of the measurements of the building cluster blocks at two different spatial scales.

[0061] In GIS-based urban building spatial distribution (such as...) Figure 4 As shown in the figure, assuming the average street width at the observation scale is W, the average characteristic length of the secondary building clusters separated by it is p times W, i.e., pW, and the number of secondary clusters forming the tertiary clusters at a larger scale is n×n, and further assuming that k of these clusters are geographically disrupted, such as water bodies, urban green spaces, mountains, etc., which are unsuitable for construction, then the characteristic length L of the tertiary building cluster is:

[0062] (2)

[0063] L represents the characteristic length of the tertiary cluster, W represents the average width of the street and natural surface (the average width of the street and other unoccupied ground within the block), p represents the average measure of the tertiary cluster (the ratio between the average measure of the blocks occupied by individual building groups separated within the tertiary building cluster and the average width W of the street), the number of secondary clusters in the tertiary cluster is n×n, and k represents the number of natural surfaces in the tertiary cluster.

[0064] The ratio of the characteristic lengths of building clusters at two different spatial scales, i.e., the ratio of the characteristic lengths of tertiary building clusters to secondary building clusters. for:

[0065] (3)

[0066] The ratio N of the measurements of the building clusters at two different spatial scales is:

[0067] (4)

[0068] According to the definition of the fractal dimension D of the urban built-up area in equation (1), it can be seen that...

[0069] (5).

[0070] The methods for determining the fractal dimension D mentioned above include:

[0071] (a) By identifying the rooftop / street GIS patches in the urban built-up area, the geometric parameters such as n, k, and p in equation (5) or their numerical fitting results in a certain area can be obtained, and the fractal dimension D of the spatial distribution of urban buildings and streets in the area can be obtained.

[0072] Compared to the homogenized underlying surface used in the original model, it still possesses the aforementioned characteristics. Figure 4 The space shown is non-uniform, but its fractal dimension D needs to be obtained through box counting.

[0073] (b) The box-counting dimension method is directly applied to the above GIS map patches to obtain the fractal dimension D value based on numerical methods that is closer to the actual physical distribution of urban buildings and conforms to the above definition.

[0074] Although there may be differences in specific numerical values ​​between the two methods mentioned above, their convergence and substitutability have been proven by relevant mathematical theories and numerical simulation studies.

[0075] Step S4: Introduce a certain proportion of asymmetric broken units on the basis of the standardized grid to simulate natural disturbances such as green space, water body, and undeveloped area, so that the urban structure model is more in line with the actual heterogeneous characteristics.

[0076] II. Construction of scale response relationship of land surface parameters.

[0077] Step S5: Divide the urban built-up area into two categories: areas occupied by buildings and areas not occupied by buildings. Based on the functional relationship between fractal dimension D and spatial scale dx, derive the scale response expressions of land surface parameters such as albedo, roughness, height, and heat capacity.

[0078] For any key land surface parameter in a surface grid within a numerical weather prediction model at any spatial scale, its value should be obtained by weighting the key land surface parameter of the natural surface and the urban built-up area contained in that grid, and using the respective area proportion of that grid as weighting coefficients, i.e.:

[0079] (6)

[0080] in, This indicates that the grid resolution is the observation scale. The average key land surface parameters within the surface grid are times higher. The value ( It also indicates the multiple between the grid resolution of the simulation area and the observation scale. This indicates the proportion of urban built-up area within the surface grid. This represents the numerical value of the key land surface parameter of the natural surface within the surface grid, both of which are typically provided by the input parameter table for numerical weather prediction. This indicates that the grid resolution on the Earth's surface is the observation scale. When the value is multiplied, the value of the same key land surface parameter in the urban built-up area at this characteristic spatial scale is derived below.

[0081] (7)

[0082] in, This indicates that the scale of interest is the observation scale. On a scale of multiples, the proportion of building roof area to the total built-up area within the urban built-up area is derived below; and These represent the values ​​of the land surface parameters at the top of the building and the top of the street valley area formed between buildings, obtained through observation (i.e., at the observation scale). These two values ​​are obtained directly from the model's input parameter table or calculated by other modules of the model and are used as input parameters for this formula.

[0083] Assuming that, on the observation scale, the roof area within the urban built-up area is... The total built-up area of ​​the city (i.e., the built-up area including both rooftop and non-rooftop areas) is Then its proportion According to the definition, it can be expressed as:

[0084] (8)

[0085] However, when the scale of interest is the observation scale... When the area of ​​a roof exhibits fractal characteristics and has a fractal dimension of D, the area of ​​the roof is multiplied by a factor of 1. It should be:

[0086] (9)

[0087] However, the built-up area of ​​a city does not exhibit fractal characteristics, therefore,

[0088] (10)

[0089] in, This indicates that when the scale of interest is the observation scale... When the area is doubled, it refers to the area of ​​the urban built-up area.

[0090] The scale effect of key land surface parameters is obtained based on the above formula:

[0091] The scale of interest is the observation scale. When the area is doubled, the ratio of roof area to built-up area is... The following relationship exists: when the shape dimension D of the urban built-up area is less than 2, the proportion of roof area increases with the characteristic scale. The decrease occurs with increasing multiples, and this decrease is determined by the fractal dimension:

[0092] (11)

[0093] That is, the influence of key land surface parameters of rooftops on the average land surface parameters of urban built-up areas decreases with the increase of characteristic scale.

[0094] Only when the fractal dimension D=2, that is, when the non-uniformity of the urban canopy space is not considered, the ratio of roof area to built-up area is independent of the spatial scale of interest. This is the result of the traditional urban canopy model assumption.

[0095] Substituting equation (11) into equation (7), and further substituting it into equation (6), we can see that:

[0096] (12)

[0097] According to equation (12), the horizontal spatial resolution of the model simulation is determined to be the observable spatial scale. When the area is multiplied by 1, the proportion of urban built-up areas to the simulated surface grid is... Average land surface parameters of natural vegetation use type Land surface parameters of rooftops within urban built-up areas Land surface parameters of man-made surfaces such as streets within urban built-up areas The input parameters of the model are provided by the measurement data based on the observation scale. The fractal dimension D of the spatial distribution of buildings in the urban built-up area is determined by equation (4) or the equivalent box-counting dimension method based on GIS data image recognition. When D < 2, this equation is consistent with D and The value of this parameter is related to the critical land surface parameter. It has a spatial scaling law. When D=2, equation (12) converges to:

[0098] (13)

[0099] In equation (13), The value of is related to the feature parameter of the scale of interest. It's irrelevant. This is consistent with traditional modeling methods.

[0100] Using albedo as a characterizing parameter, the spatial scaling law variation of this parameter in the urban canopy underlying surface with fractal characteristics is analyzed at a 1km observation scale.

[0101] The scale variation characteristics of the albedo parameter in the spatially heterogeneous fractal urban canopy model are compared with those in the traditional urban canopy model to clarify the differences between the two in the spatiotemporal variation of the average albedo within the urban surface grid.

[0102] Traditional urban canopy models typically assume a fixed surface albedo value that does not change with the urban spatial scale. However, in the spatially non-homogeneous fractal urban canopy model used in this invention, the proportion of non-building areas (such as green spaces, bare soil, and water bodies) increases with the urban scale. Since the albedo of these non-building surfaces is generally lower than that of building surfaces, this leads to a decrease in the overall average albedo of the urban canopy. Model validation results confirm that this fractal model successfully captures and reproduces this physical process of albedo variation with the urban scale.

[0103] Third, a two-way physical coupling is performed between the spatially heterogeneous fractal urban canopy model and the mesoscale numerical weather model.

[0104] Step S6: Couple the above spatially non-uniform fractal urban canopy model with the mesoscale numerical weather model in two directions, so that the land surface parameters can be adaptively adjusted with the grid scale.

[0105] This paper embeds a spatially heterogeneous fractal urban canopy model into a mesoscale numerical weather prediction model. By bidirectionally coupling the land-atmosphere water and heat flux transport processes, cross-scale optimization simulation of a nested grid system is achieved, using the WRF (Weather Research and Forecasting) model as an example. This includes:

[0106] The spatially heterogeneous fractal urban canopy model is embedded into the Noah Land Surface Process Model (LSM) Urban Submodule (UCM) of the WRF model, replacing the homogenized underlying surface parameter scheme;

[0107] Urban morphological parameters such as planar index and windward area index are dynamically calculated based on grid resolution.

[0108] At each time step, the system receives near-surface temperature, humidity, and wind speed fields output by the WRF, and generates variables such as sensible heat, latent heat, and momentum turbulence flux, which are then fed back to the atmospheric module.

[0109] The spatial non-uniform geometric model of urban building space and the spatial scaling law of land surface parameters are coupled to the multi-level nested grid system of the mesoscale numerical weather prediction model through the spatial non-uniform fractal urban canopy model module. The values ​​of their respective key land surface parameters are determined according to their different spatial resolutions, so as to fully consider the spatial non-uniform geometric model of urban building distribution and the spatial scaling law of urban land surface parameters in the prediction model system.

[0110] Within the framework of the aforementioned two-way physical coupling, different land cover types significantly influence atmospheric circulation, near-surface temperature, and convective activity through energy and moisture exchange with the atmosphere, thereby altering regional precipitation patterns. Compared to first-order variables such as temperature and wind speed, which are dominated by boundary conditions and exhibit relatively gentle spatial variations, precipitation is a high-order feedback variable. Its simulation heavily relies on a high-precision description of the three-dimensional spatial distribution of meteorological elements, and simulation errors are more likely to expose structural defects in the model's parameterization scheme. Therefore, using precipitation simulation results can sensitively and comprehensively evaluate the response capability of two-way coupled models to the spatial non-uniformity of meteorological elements in urban areas.

[0111] Step S7: Dynamically call the land surface parameter module in the urban meteorological model (this invention takes WRF as an example, but CMAQ, the air quality forecast and assessment system, can also be used), and use the updated numerical weather model coupled with the spatially heterogeneous fractal urban canopy model to calculate heat, momentum, and water flux.

[0112] Step S8: Configure the initial and boundary conditions of the configuration mode, including setting the building boundary data source, inputting the fractal dimension D value, defining the ratio of the observation scale to the simulation scale, and completing the grid resolution settings (in this case, three nested layers are selected, with grid resolutions of 32km-8km-2km respectively).

[0113] IV. Numerical Experiments and Effect Evaluation.

[0114] Step S9: Select a typical urban area for extreme heavy rainfall events and simulate them using both a traditional urban canopy model and the improved model of this invention. The simulation results for the experimental and control groups are as follows: Figure 8 As shown.

[0115] We selected typical urban areas for severe convective precipitation events, conducted sensitivity numerical simulation and forecast accuracy verification experiments, and evaluated the improvement effect of the above models.

[0116] Based on real-time meteorological boundary conditions and urban structure data, an improved model containing a fractal urban canopy model and a traditional homogenization model are respectively driven to simulate a typical urban heavy precipitation process.

[0117] An experimental group (using an improved spatially non-uniform fractal urban canopy model) and a control group (using the original model of a traditional homogenized urban canopy model) were set up to compare and analyze the simulation results of the two models for the same typical heavy precipitation event. The results were comprehensively evaluated from multiple dimensions such as precipitation intensity, spatiotemporal evolution, and spatial distribution accuracy.

[0118] The accuracy and applicability of simulated precipitation results are measured by two indicators: the relative error (BIAS) of simulated precipitation data compared to observed precipitation data and the correlation coefficient (r). The calculation formulas are defined as follows:

[0119] (14)

[0120] (15)

[0121] Where BIAS represents the relative error, and r represents the correlation coefficient. and These represent forecasted precipitation and observed precipitation, respectively. This represents the average value of the forecast precipitation. This represents the average observed precipitation, and n represents the number of data points.

[0122] Step S10: Compare and analyze the simulation results with the actual observation data to verify the model improvement effect in terms of precipitation spatial distribution, intensity extremes, and temporal evolution. Experiments show that after introducing a spatially non-uniform fractal urban canopy model, the model more accurately captures the spatial location and development process of the rainstorm core area.

[0123] This invention characterizes the non-uniformity of urban building spatial distribution through fractal dimension D, and further discovers the relationship between the numerical values ​​of key land surface parameters and model resolution, i.e., the spatial scale of the features of interest—that is, the spatial scaling law characteristics of urban land surface parameters. This invention is applicable to the improved simulation of urban land surface processes in atmospheric boundary layer numerical models, enabling dynamic adjustment of land surface parameters with spatial resolution. This model can be embedded into the urban canopy module of meteorological models, improving the spatiotemporal resolution and accuracy of the model for forecasting heavy urban precipitation.

[0124] Example 2

[0125] Based on the precipitation forecasting method based on a spatially heterogeneous fractal urban canopy model described in Embodiment 1, this embodiment provides a precipitation forecasting system based on a spatially heterogeneous fractal urban canopy model, including: a data acquisition module for collecting meteorological data of a target urban area; and a precipitation forecasting module for inputting the collected meteorological data into a mesoscale numerical weather prediction model that integrates the spatially heterogeneous fractal urban canopy model, and outputting precipitation forecast results.

[0126] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A precipitation forecasting method based on a spatially heterogeneous fractal urban canopy model, characterized in that, include: Collect meteorological data for the target urban area; The collected meteorological data is input into a mesoscale numerical weather prediction model that incorporates a spatially heterogeneous fractal urban canopy model, and the precipitation forecast results are output. The methods for constructing spatially non-uniform fractal urban canopy models include: Acquire remote sensing images or GIS building boundary data of the target urban area and perform gridding to obtain a surface grid. At the same time, divide the surface grid into urban built-up areas and natural surfaces. Obtain key land surface parameters of urban built-up areas at the observation scale, and derive the spatial scaling law characteristics of the key land surface parameters; Based on fractal geometry methods, the fractal dimension of the target urban area is calculated. Based on the spatial scaling law characteristics of fractal dimension and key land surface parameters, a spatially non-uniform fractal urban canopy model is established. The urban built-up area is composed of multi-level self-similar building clusters, including: A "community" is formed by a number of buildings and the streets between them, i.e., a primary cluster. A "community" is composed of multiple "cells" and the streets between them, i.e., a secondary cluster. Urban areas formed by multiple "communities" and the streets or gaps between "communities" are called three-level clusters; Multiple secondary clusters, tertiary clusters, and roads between clusters constitute the urban built-up area; Based on fractal geometry methods, the fractal dimension of the target urban area is calculated, specifically: The fractal dimension D of the urban built-up area is defined as: ； in, This represents the ratio of the feature lengths of the tertiary cluster block to the secondary cluster block, where N represents the ratio of the measurements of the building cluster blocks at two different spatial scales. ； ； ； L represents the characteristic length of the tertiary cluster, W represents the average width of the street and natural surface, p represents the average measure of the tertiary cluster, the number of secondary clusters in the tertiary cluster is n×n, and k represents the number of natural surfaces in the tertiary cluster. The value of any key land surface parameter in a given surface grid is obtained by weighting the key land surface parameters of both the natural surface and the urban built-up area contained in that surface grid, as well as their respective area proportions within that surface grid, using weighting coefficients. ； in, This indicates that the grid resolution is the observation scale. The average key land surface parameters within the surface grid are times higher. The value; This indicates the proportion of urban built-up area within the surface grid. This represents the numerical value of the key land surface parameter of the natural surface within the grid. This indicates that the grid resolution on the Earth's surface is the observation scale. When the resolution is multiplied, the value of the same key land surface parameter in the urban built-up area at the feature space scale corresponding to the current grid resolution is obtained.

2. The precipitation forecasting method according to claim 1, characterized in that, The surface grid is divided into urban built-up areas and natural surfaces, including: The urban built-up area is divided into the rooftop area and the street area; Numerical statistical analysis was performed on the spatial layout characteristics of building rooftop and street areas based on patch recognition technology, and the statistical analysis results were obtained. The statistical analysis results were binarized, and the building rooftop coverage area and street area were marked separately. The street area includes: roads and squares; the natural surface includes: natural features within the urban built-up area, including vegetation-covered areas, water bodies and non-buildable mountainous areas.

3. The precipitation forecasting method according to claim 1, characterized in that, The fusion of spatially heterogeneous fractal urban canopy models with mesoscale numerical weather prediction models is as follows: By embedding a spatially heterogeneous fractal urban canopy model into a mesoscale numerical weather prediction model, cross-scale optimization simulation of a nested grid system is achieved through bidirectional coupling of water and heat flux transport processes between the Earth and the atmosphere, including: The spatially heterogeneous fractal urban canopy model is embedded into the urban submodule of the land surface process model in the mesoscale numerical weather prediction model, replacing the existing underlying surface parameter scheme. Urban morphological parameters, including planar index and windward area index, are dynamically calculated based on grid resolution. At each time step, the system receives near-surface temperature, humidity, and wind speed fields output by the mesoscale numerical weather prediction model, generates sensible heat, latent heat, and momentum turbulence, and feeds them back to the atmospheric module.

4. The precipitation forecasting method according to claim 1, characterized in that, The mesoscale numerical weather prediction models include the WRF model and the CMAQ model.

5. The precipitation forecasting method according to claim 1, characterized in that, The accuracy of precipitation forecasts is quantitatively assessed by calculating relative error and correlation coefficient: ； ； Where BIAS represents the relative error, and r represents the correlation coefficient. and These represent forecasted precipitation and observed precipitation, respectively. This represents the average value of the forecast precipitation. This represents the average observed precipitation, and n represents the number of data points.

6. A precipitation forecasting system based on a spatially heterogeneous fractal urban canopy model, characterized in that, include: The data acquisition module is used to collect meteorological data for the target urban area; The precipitation forecast module is used to input the collected meteorological data into a mesoscale numerical weather prediction model that incorporates a spatially heterogeneous fractal urban canopy model, and output precipitation forecast results. The methods for constructing spatially non-uniform fractal urban canopy models include: Acquire remote sensing images or GIS building boundary data of the target urban area and perform gridding to obtain a surface grid. At the same time, divide the surface grid into urban built-up areas and natural surfaces. Obtain key land surface parameters of urban built-up areas at the observation scale, and derive the spatial scaling law characteristics of the key land surface parameters; Based on fractal geometry methods, the fractal dimension of the target urban area is calculated. Based on the spatial scaling law characteristics of fractal dimension and key land surface parameters, a spatially non-uniform fractal urban canopy model is established. The urban built-up area is composed of multi-level self-similar building clusters, including: A "community" is formed by a number of buildings and the streets between them, i.e., a primary cluster. A "community" is composed of multiple "cells" and the streets between them, i.e., a secondary cluster. Urban areas formed by multiple "communities" and the streets or gaps between "communities" are called three-level clusters; Multiple secondary clusters, tertiary clusters, and roads between clusters constitute the urban built-up area; Based on fractal geometry methods, the fractal dimension of the target urban area is calculated, specifically: The fractal dimension D of the urban built-up area is defined as: ； in, This represents the ratio of the feature lengths of the tertiary cluster block to the secondary cluster block, where N represents the ratio of the measurements of the building cluster blocks at two different spatial scales. ； ； ； L represents the characteristic length of the tertiary cluster, W represents the average width of the street and natural surface, p represents the average measure of the tertiary cluster, the number of secondary clusters in the tertiary cluster is n×n, and k represents the number of natural surfaces in the tertiary cluster. The value of any key land surface parameter in a given surface grid is obtained by weighting the key land surface parameters of both the natural surface and the urban built-up area contained in that surface grid, as well as their respective area proportions within that surface grid, using weighting coefficients. ； in, This indicates that the grid resolution is the observation scale. The average key land surface parameters within the surface grid are times higher. The value; This indicates the proportion of urban built-up area within the surface grid. This represents the numerical value of the key land surface parameter of the natural surface within the grid. This indicates that the grid resolution on the Earth's surface is the observation scale. When the resolution is multiplied, the value of the same key land surface parameter in the urban built-up area at the feature space scale corresponding to the current grid resolution is obtained.

Images