Urban underlying surface multi-dimensional index system construction method

Abstract

The invention discloses an urban underlying surface multi-dimensional index system construction method, which comprises the following steps of S1, collecting data, constructing an underlying surface database with high temporal-spatial resolution, and obtaining year-by-year LUCC data in a research time period; s2, dividing sub-basin grids according to long-time-sequence year-by-year LUCC data, determining a threshold value of a plaque area, analyzing underlying surface space-time evolution characteristics in a given grid, and constructing a representation index; s3, researching space-time evolution characteristics of multiple dimension indexes, and constructing an underlying surface characteristic index alternative library; s4, constructing a multi-dimensional representation index system of the underlying surface; and S5, processing data in the underlying surface multi-dimensional characterization index system by using a data dimension reduction technology, and constructing an underlying surface characterization index. According to the method, through a more comprehensive and more accurate representation system, the attribute, geometric and spatial distribution characteristics of a complex underlying surface are completely expressed, and researchers and managers are helped to better deal with a drainage basin comprehensive management problem in a changing environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of urban underlying surface representation, and in particular relates to a method for constructing a multi-dimensional indicator system for urban underlying surfaces. Background Art

[0002] The change of the underlying surface is one of the basic driving factors for the evolution of the earth's various spheres, and its spatial pattern is mainly affected by the topography and the distribution of land use patches. In the 1990s, the LUCC study jointly initiated by IGBP and IHDP used area as the main characteristic indicator and adopted methods such as field experiments and statistical analysis to analyze the changing laws of LUCC.

[0003] Entering the 21st century, with the intensification of human activities, the spatial pattern characteristics of the underlying surface have received more attention. Researchers have proposed a variety of quantitative analysis methods, such as remote sensing technology combined with GIS spatial analysis, landscape index evaluation, spatial statistics, and model simulation, to study the spatial pattern characteristics of the underlying surface and its environmental impact, revealing the comprehensive effects of land use change, urbanization, ecosystem services, and climate change on surface processes.

[0004] However, previous studies often only focus on a single or a few indicators, which may ignore the intrinsic connection between the underlying surface characteristics and their compound effects on environmental processes, causing problems in the understanding of watershed runoff and integrated watershed management. Therefore, it is necessary to carry out multi-dimensional characterization and method research on complex underlying surfaces. The present invention comprehensively considers the underlying surface characteristics from three dimensions: attribute characteristics, geometric characteristics, and spatiotemporal distribution changes. It can more accurately understand and predict the potential impact of underlying surface changes on the environment, and provide a scientific basis for watershed runoff, urban planning, and environmental management. Summary of the invention

[0005] In order to solve the above problems, the present invention proposes a method for constructing a multi-dimensional indicator system for urban underlying surfaces.

[0006] The method for constructing a multi-dimensional indicator system for urban underlying surfaces of the present invention comprises the following steps:

[0007] S1. Collect data, build a high-temporal and spatial resolution underlying surface database, and obtain annual LUCC data during the study period: For complex underlying surfaces in watersheds containing urban areas, conduct research on multi-source remote sensing image data fusion algorithms; use the fused high-precision long-term remote sensing images to extract and analyze base year land use coverage / type (LUCC) data based on deep learning; collect DEM, precipitation, road distribution, population distribution and other data, establish a natural-social environment auxiliary data set, and study the construction method of the dynamic development matrix of the underlying surface spatial pattern; based on the base year LUCC data and the dynamic development matrix, study the interpolation and reconstruction algorithm of the land use data in the missing years, and obtain annual LUCC data during the study period;

[0008] S2. Based on the long-term annual LUCC data, sub-basin grids are divided, and the threshold of patch area is determined. The spatiotemporal evolution characteristics of the underlying surface in a given grid are analyzed, and characterization indicators are constructed: Combined with high-precision long-term annual LUCC data and DEM data, sub-basin grid division and patch area threshold determination methods are studied; the spatiotemporal transformation trajectory of each land use patch in a given grid is analyzed, and the spatiotemporal evolution characteristics of the underlying surface are studied; from multiple dimensions such as "attribute characteristics", "geometric characteristics" and "spatial distribution" of the underlying surface patches, a comprehensive and accurate characterization method of the spatial pattern of the complex underlying surface is studied;

[0009] S3. Study the spatiotemporal evolution characteristics of multiple dimensional indicators and build a candidate library of underlying surface characteristic indicators: Based on the LUCC data of the sub-basin, analyze the scale effect, set an appropriate patch area threshold, and use a combination of landscape pattern analysis (LSA), morphological spatial pattern analysis (MSPA) and spatial trajectory analysis to study the spatiotemporal evolution characteristics of multi-dimensional indicators such as underlying surface patch type, terrain index, NDVI index, NDBI index, "attribute characteristics" (ATtribute), "geometric characteristics" (GEometric) such as area and shape, and "spatial distribution" (DIstribution) such as connectivity and dispersion, and build a candidate library of underlying surface characteristic indicators;

[0010] S4. Construct a multi-dimensional characterization index system for the underlying surface: On the basis of completing data collection and analysis of spatiotemporal evolution characteristics, construct a comprehensive multi-dimensional characterization index system for the underlying surface; this system will cover key dimensions such as attribute characteristics, geometric characteristics, and spatial distribution, and evaluate the complexity of underlying surface characteristics through quantitative methods; specific operations include: determining key indicators for each dimension, assigning appropriate weights to these indicators, and developing a comprehensive scoring system that can reflect the contribution of underlying surface characteristics in different dimensions;

[0011] S5. Use data dimensionality reduction technology to process the data in the underlying surface multi-dimensional characterization index system and construct the underlying surface characterization index: After establishing the multi-dimensional characterization index system, use data dimensionality reduction technology, such as principal component analysis (PCA) and linear discriminant analysis (LDA), to process the data in the index system; the purpose is to identify the most representative and independent key indicators from a large number of original indicators. These indicators can reflect the information of the underlying surface characteristics to the greatest extent while reducing the complexity of the data; through the evaluation of correlation and independence, the most critical indicators for the characterization of the underlying surface characteristics are selected, and finally a multi-dimensional characterization index system for complex underlying surfaces is constructed; at the same time, the hierarchical analysis method is used to calculate an indicator that comprehensively reflects the characteristics of the complex underlying surface to simplify the expression of the underlying surface situation.

[0012] The S1 obtains the annual LUCC data during the study period. The specific process is to fuse the multi-source remote sensing image data of the complex underlying surface of the watershed containing the urban area; use the fused high-precision long-time series remote sensing images to extract and calculate the land use coverage / type (LUCC) data; then establish a natural-social environment auxiliary data set to study the dynamic development matrix of the underlying surface spatial pattern, and finally obtain the annual LUCC data during the study period.

[0013] The multiple dimensional indicators in S3 include attribute features, geometric features and spatial distribution.

[0014] The steps of constructing the underlying surface multi-dimensional characterization index system in S4 are: determining the key indicators of each dimensional indicator, assigning appropriate weights to these key indicators, and developing a comprehensive scoring system.

[0015] The key characteristic indicators of each dimensional indicator in S4 are: attribute characteristics include terrain attributes, soil attributes, vegetation attributes, physical attributes and surface cover attributes; geometric characteristics include the area of ​​various land use types, the number of patches, the perimeter of patches and the complexity of patch shapes; spatial distribution includes connectivity, dispersion, aggregation and diversity.

[0016] The data in the multi-dimensional characterization index system of the underlying surface is processed by using data dimensionality reduction technology in S5. The purpose is to identify the most representative and independent key indicators from a large number of original indicators through the evaluation of correlation and independence, so as to provide a solid data foundation for the final constructed characterization index.

[0017] The step of S5 comprises:

[0018] S501. Data collection and preprocessing;

[0019] S502. Data standardization;

[0020] S503. Correlation analysis;

[0021] S504. Data dimensionality reduction processing;

[0022] S505. Dimensionality reduction screening and analysis of feature indicators.

[0023] The indicators in the underlying surface multi-dimensional characterization index system are divided into primary indicators, secondary indicators and tertiary indicators. The primary indicators refer to attribute characteristics, geometric properties and spatial distribution. Each primary indicator has corresponding multiple secondary indicators, and each secondary indicator also has corresponding multiple tertiary indicators.

[0024] The underlying surface characterization index includes the primary index, the secondary index and the underlying surface comprehensive index.

[0025] The three first-level indexes are expressed as follows:

[0026] Underlying surface properties:

[0027] Among them, A T Represents the overall score of the underlying surface attribute characteristics, n T is the number of secondary indicators under the attribute characteristics, w Ti is the weight of the secondary index i in the attribute feature, S Ti is the standardized value of the secondary indicator i;

[0028] Geometric characteristics of underlying surface:

[0029] Among them, A G Represents the overall score of the underlying surface geometric characteristics, n G is the number of secondary indices under geometric characteristics, w Gj is the weight of the secondary index j in the geometric characteristics, S Gj is the standardized value of the secondary indicator j;

[0030] Spatial distribution of underlying surface:

[0031] Among them, A D Represents the overall score of the underlying surface geometric characteristics, n D is the number of secondary indicators under spatial distribution, w Dk is the weight of the secondary index k in the spatial distribution, S Dk is the standardized value of the secondary indicator k.

[0032] The scores of the three first-level indicators are integrated to obtain a comprehensive underlying surface index:

[0033] CIS=ω A ×AT+ω G ×GE+ωD ×DI

[0034] Among them, CIS represents the comprehensive index of the underlying surface of the basin in the study area, which comprehensively reflects the characteristics of the attribute characteristics, geometric characteristics, and spatial distribution in the basin; AT, GE, and DI are the values ​​of the three first-level indicators of underlying surface attribute characteristics, geometric characteristics, and spatial distribution, respectively. A ,ω G and ω D They are the weights of underlying surface attribute characteristics, geometric properties and spatial distribution.

[0035] The beneficial effects of the present invention are:

[0036] The present invention starts with analyzing the spatiotemporal evolution characteristics of the complex underlying surface in the watershed, and characterizes the characteristics of the complex underlying surface through three dimensions: "attribute characteristics", "geometric characteristics" and "spatial distribution". It is not limited to the traditional area characterization and the characterization of relative spatial relationships. The present invention will fully express the attributes, geometry and spatial distribution characteristics of the complex underlying surface through a more comprehensive and accurate characterization system, which will help researchers and managers to better cope with the issues of comprehensive watershed management under a changing environment. DETAILED DESCRIPTION

[0037] This application conducts research on multi-source remote sensing image data fusion algorithms for complex underlying surfaces in watersheds that include urban areas, and constructs an underlying surface database with high temporal and spatial resolution; analyzes the spatiotemporal evolution characteristics of the underlying surface, comprehensively and accurately classifies the spatial pattern of the underlying surface according to different dimensions, and constructs characterization indicators; analyzes the correlation between variables, and screens out relatively independent underlying surface characteristic indicators.

[0038] S1. Fusion of multi-source remote sensing image data of complex underlying surfaces in watersheds including urban areas; extraction and calculation of land use coverage / type (LUCC) data using fused high-precision long-term remote sensing images.

[0039] Based on high-precision long-term LUCC data, combined with DEM data, precipitation, road distribution, population distribution and other data in the study area, a natural-social environment auxiliary data set is established. In particular, due to the diversity of data sources, there are inconsistencies in the format, spatiotemporal resolution (accuracy) and other aspects of different data. Therefore, when conducting analysis, we must first determine the basic unit of the study (including time and space) and unify the multi-source data into the determined basic unit. Specifically:

[0040] (1) Remote sensing image data

[0041] The high-resolution series (GF1-GF4) satellite data in the study area were collected, and the long-term remote sensing images (LandSat series, MODIS series) collected by Pixel Information ExpertEngine (PIE-Engine) or Google Earth Engine (GEE) platform were corrected and fused. For the image data that met the requirements, the Classification and Regression Tree (CART), Random Forest (RF) and other algorithms were used to extract the land use type / cover data and obtain the land use / cover (LUCC) data set.

[0042] The long-term high-resolution precipitation and evaporation datasets (such as ERA-5, GPM, etc.), road datasets (OpenGMS), digital elevation data (DEM), and social environment auxiliary datasets such as population datasets were collected in the study area; based on the obtained benchmark year LUCC dataset, the Ordinary Least Squares residuals-based MOving SUM (OLS-MOSUM) algorithm was used to identify the changes between pixels and construct the dynamic development matrix of LUCC suitability in each benchmark year. Based on the dynamic development matrix, the linear regression (LR) and convolutional neural network algorithm (CNN) were used to establish an interpolation and reconstruction algorithm for land use data in missing years to obtain the annual LUCC data during the study period.

[0043] (2) Precipitation and runoff data

[0044] The data in the collected precipitation raster dataset are calibrated using the daily precipitation data from the measured sites, and the precipitation data in NC format are organized into binary array format data that is consistent with the runoff data.

[0045] (3) DEM elevation data

[0046] The collected DEM data were combined with the measured topography and river network data to divide the sub-basins using PyFlwDir. Taking into account the area, socioeconomic status and other factors of each sub-basin, the sub-basins were appropriately merged to form the basic research unit, so that the division of the basic unit took into account both ecological significance and statistical rationality.

[0047] (4) Social Factors Dataset

[0048] The collected transportation network data such as railways and roads are processed into raster data using the Euclidean distance tool; the population density data are processed by removing null values ​​using the raster calculator.

[0049] S2. Combine high-precision long-term annual LUCC data and DEM data to study the sub-basin grid division and patch area threshold determination methods; analyze the spatiotemporal transformation trajectories of each land use patch within a given grid, and study the spatiotemporal evolution characteristics of the underlying surface; from multiple dimensions such as "attribute characteristics", "geometric characteristics" and "spatial distribution" of the underlying surface patches, study the comprehensive and accurate characterization methods of the complex underlying surface spatial pattern.

[0050] S3. Based on the LUCC data of the sub-basin, the scale effect is analyzed, and the appropriate patch area threshold is set. The "attribute characteristics" such as the underlying surface patch type, terrain index, NDVI index, NDBI index, "geometric characteristics" such as area and shape, and the spatiotemporal evolution characteristics of multi-dimensional indicators such as "spatial distribution" such as connectivity and dispersion are studied by combining landscape pattern analysis, morphological spatial pattern analysis and spatial trajectory analysis. A candidate library of underlying surface characteristic indicators is also constructed.

[0051] S4. On the basis of completing data collection and analysis of spatiotemporal evolution characteristics, a comprehensive underlying surface multi-dimensional characterization index system is constructed. This system will cover key dimensions such as attribute characteristics, geometric characteristics, and spatial distribution, and evaluate the complexity of underlying surface characteristics through quantitative methods. Specific operations include: determining the key indicators of each dimension, assigning appropriate weights to these indicators, and developing a comprehensive scoring system that can reflect the contribution of underlying surface characteristics in different dimensions. The specific calculation formula for the characterization index that can be used for quantification is as follows:

[0052] 1. Attribute characteristics

[0053] (1) Terrain attributes

[0054] 1) Ruggedness: The standard deviation of terrain elevation changes, reflecting the severity of terrain changes.

[0055]

[0056] 2) Slope:

[0057]

[0058] (2) Soil properties

[0059] 1) Soil fertility: Soil fertility, pH value, organic matter content, etc. can be determined through laboratory testing of soil samples.

[0060] 2) Soil Moisture Content:

[0061]

[0062] 3) Soil permeability (Infiltration Rate):

[0063] f=K×(P-Suction)

[0064] Where K is the saturated hydraulic conductivity of the soil, P is the applied pressure, and Suction is the suction of the soil.

[0065] (3) Vegetation attributes

[0066] 1) Vegetation Cover:

[0067]

[0068] 2) Leaf Area Index (LAI):

[0069]

[0070] 3) Vegetation index (such as normalized difference vegetation index, NDVI):

[0071]

[0072] Among them, NIR is the reflectivity of the near-infrared band, and R is the reflectivity of the visible red band.

[0073] (4) Physical properties

[0074] 1) Terrain Roughness Index (TRI): A commonly used method to quantify surface roughness, which can be calculated using the following formula:

[0075]

[0076] Among them, the standard deviation (SD) is the standard deviation of the deviation between the elevation value of each point on the DEM and its average elevation value, which can be used to measure the roughness of the surface.

[0077] 2) Heat capacity refers to the amount of heat absorbed or released per unit mass of a substance within a unit temperature range. In some cases, empirical models or formulas can be used to estimate the surface heat capacity. For example, the following simplified formula can be used to estimate the heat capacity of soil:

[0078] C=(1-ω)C solid +ωC water

[0079] Where C is the heat capacity of the soil, ω is the moisture content of the soil, and C solidis the heat capacity of the soil solid phase, C water is the heat capacity of water.

[0080] (5) Land cover properties

[0081] The relationship between artificial surfaces and natural surfaces in a watershed can be expressed by the impervious surface area ratio:

[0082]

[0083] 2. Geometric features

[0084] (1) Area, number of patches, and perimeter of patches for each land use type.

[0085] Spatial analysis tools of GIS software can be used to calculate the number, area, perimeter, etc. of patches.

[0086] (2) Complexity of plaque shape

[0087] 1) Perimeter-Area Fractal Dimension (PAFRAC):

[0088]

[0089] Where P is the perimeter of the patch and A is the area of ​​the patch.

[0090] 2) Landscape Shape Index (LSI):

[0091] LSI with square as reference:

[0092]

[0093] Among them, E represents the total length of all patch boundaries in the landscape, and A represents the total area of ​​the landscape. This index is used to measure the irregularity of patch shapes. The larger the value, the more complex the patch shape and the more irregular it tends to be.

[0094] LSI using a circle as a reference:

[0095]

[0096] The meanings of E and A are the same as above, but a circle is used as a reference to calculate the degree of deviation of the patch shape from the circle.

[0097] 3) Area-weighted mean shape index (AWMSI), calculated as follows:

[0098]

[0099] Among them, A i is the area of ​​the ith patch, LSI i is the shape index of the ith patch (usually LSI, landscape shape index), A total is the total area of ​​the landscape.

[0100] 4) Shape Complexity Index (SCI), the calculation formula can be expressed as:

[0101]

[0102] Where E is the actual perimeter of the patch, P min is the minimum perimeter of a regular shape (usually a circle or a square) with the same area. For a circle, For a square, Where A is the area of ​​the patch.

[0103] 3. Spatial distribution characteristics

[0104] (1) Connectivity

[0105] 1) Integral index of connectivity (IIC):

[0106]

[0107] Where n is the total number of patches in the landscape, a i and a j are the areas of patch i and patch j respectively, nl ij is the number of connections between plaque i and plaque j, A L is the area of ​​the background landscape, 0≤IIC≤l, IIC of 0 means there is no connection between habitat patches, and IIC of 1 means the entire landscape is a habitat patch.

[0108] 2) Probability of connectivity (PC):

[0109]

[0110] Among them, P ij * is the maximum possibility of direct diffusion of species between patches i and j, 0<PC<1.

[0111] 4) Importance Value of Patches (dPC): The importance of patches in maintaining landscape connectivity is measured by calculating the impact of patch removal on the overall connectivity index:

[0112] dPC=(PC-PC remove ) / PC×100%

[0113] Among them, PC remove It is the overall index value of the remaining patches after removing a single patch. dPC refers to the change in PC after a patch is removed to measure the importance of the patch for maintaining landscape connectivity.

[0114] 5) Connectivity (CONNECT):

[0115]

[0116] Among them, c ijk is the connection status of patches j and k related to patch type i within the critical distance, n i is the number of patches of patch type i in the landscape, and connectivity is equal to the number of nodes between all patches of a certain type (when patches j and k are connected, c ijk =1; otherwise, c ijk =0) divided by the number of all possible nodes. The range of the connectivity index is [0, 100]. The higher the connectivity of the patch, the larger the connectivity index.

[0117] 6) The contagion index (CONTAG) is an important landscape ecology indicator used to measure the aggregation degree or extension trend of different patch types in the landscape.

[0118]

[0119] Where Pi is the percentage of area occupied by type i patches; gik is the number of adjacent type i patches and type k patches; and m is the total number of patch types in the landscape.

[0120] 7) Spatial Autocorrelation Index: It measures the degree of spatial connection between similar values ​​in spatial data, which can indirectly reflect spatial diversity. In global correlation analysis, the most commonly used statistic is Global Moran'I, which is mainly used to describe the average degree of correlation between all spatial units and surrounding areas in the entire region. The calculation formula is as follows:

[0121]

[0122] in, n is the total number of spatial units, y i and j Respectively represent the attribute values ​​of the i-th spatial unit and the j-th spatial unit, is the mean value of all spatial unit attribute values, w ij is the spatial weight value.

[0123] (2) Discreteness

[0124] 1) Patch density (PD): The number of patches per unit area, reflecting the degree of landscape fragmentation.

[0125] 2) Edge density is an indicator used to measure the ratio of the total length of patch edges in a landscape to the total area of ​​the landscape. It reflects the frequency of contact between different patch types in the landscape. Edge density can be expressed by the following formula:

[0126]

[0127] Here, “total edge length” refers to the cumulative length of all patch edges in the landscape, while “total landscape area” refers to the total area of ​​the study area.

[0128] 3) Landscape Separation Index (Si): This index is used to measure the degree of separation between patches of the same type in the landscape. The larger the value, the higher the degree of separation between patches. The calculation formula is:

[0129]

[0130] Where A is the total area of ​​the study area, A i is the patch area of ​​landscape type i.

[0131] (3) Aggregation

[0132] 1) Cohesion Index is an indicator used in landscape ecology to measure landscape connectivity, which helps to understand the degree of connection between different landscape fragments. The calculation formula of the cohesion index is as follows:

[0133]

[0134] Among them, a ij represents the area of ​​the jth patch in the i-th type of landscape, P ij represents the perimeter of the jth patch in the i-th type of landscape, A is the area of ​​the entire landscape, and n is the number of patches in landscape type i.

[0135] 2) LandscapeAggregation Index (AI): The aggregation index is measured by calculating the degree of connectivity within the patch type, which takes into account the compactness of the distribution of patches within the same landscape type.

[0136]

[0137] Where n is the total number of patches of a certain type in the landscape, N i It is the number of patches of the same type that share a common boundary with the ith patch. The AI ​​value usually ranges from -1 to 1. Positive values ​​indicate that patches tend to be clustered, negative values ​​indicate that patches are more dispersed, and values ​​close to 0 indicate that patches are randomly distributed.

[0138] (4) Diversity

[0139] 1) Shannon's Diversity Index (SHDI):

[0140]

[0141] Among them, Pi is the proportion of the area of ​​the i-th type of patch in the landscape to the total area, n is the total number of patch types, and the value of the Shannon diversity index ranges from 0 to infinity. The larger the value, the higher the diversity of the system. When all types in the landscape appear in the same proportion, the diversity index reaches its maximum value. If one type dominates the landscape and other types are rare or missing, then the diversity index will decrease.

[0142] 2) Shannon's Evenness Index (SHEI):

[0143]

[0144] Among them, SHEI is the Shannon evenness index, SHDI is the Shannon diversity index, S is the total number of different patch types in the landscape, and the value of the Shannon evenness index is between 0 and 1. The closer the value is to 1, the more evenly the distribution of patch types in the landscape is, and the closer the value is to 0, the more uneven the distribution is. If all patch types are evenly distributed, that is, each type of patch occupies the same proportion of area, SHEI will reach the maximum value of 1; if a certain patch type dominates and other types are rare or non-existent, SHEI will be close to 0.

[0145] S5. After establishing a database of candidate multi-dimensional characterization indicators, data dimensionality reduction techniques such as principal component analysis (PCA) and linear discriminant analysis (LDA) are used to process the data in the indicator system. The purpose is to identify the most representative and independent key indicators from a large number of original indicators. These indicators can reflect the information of the underlying surface characteristics to the greatest extent while reducing the complexity of the data. By evaluating the correlation and independence, the indicators that are most critical to the characterization of the underlying surface characteristics are selected to provide a solid data foundation for the final constructed characterization index.

[0146] Specifically, you can follow the steps below:

[0147] (1) Data collection and preprocessing: Data cleaning is performed on all underlying surface characteristic indicators to deal with missing values ​​and outliers.

[0148] (2) Data standardization: Since methods such as PCA are affected by the data scale, the data needs to be standardized so that each indicator has a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0149] (3) Correlation analysis: Before dimensionality reduction, correlation analysis is performed, such as calculating the Pearson correlation coefficient matrix to identify highly correlated variable pairs.

[0150] (4) Data dimensionality reduction processing:

[0151] 1) Principal Component Analysis (PCA): Apply PCA to reduce the dimensionality of a dataset while retaining most of the variability in the dataset. Interpret the principal components to see which original variables are highly correlated with the principal components.

[0152] 2) Linear Discriminant Analysis (LDA): If the data set has multiple categories, LDA can be used to find the best dimensionality reduction direction that not only retains the variability of the data but also takes into account the separability of the categories. LDA generates discriminant coefficients that can be used to select the most discriminative feature indicators.

[0153] 3) Local Linear Embedding (LLE): LLE is a nonlinear dimensionality reduction method based on local preservation of neighboring points. LLE is applied to discover the intrinsic structure of data and identify relatively independent variables in low-dimensional space.

[0154] (5) Feature index dimensionality reduction screening and analysis: Analyze the dimensionality reduction results of methods such as PCA, LDA, and LLE to see which original variables maintain a high degree of independence in the space after dimensionality reduction.

[0155] 1) Feature screening: Based on the results of the dimensionality reduction method, select those variables that have a large contribution and are independent of each other in the dimensionality reduction space. Consider the explanatory power, stability, and relevance of the variables to the research purpose.

[0156] 2) Model validation: Use leave-one-out cross-validation (LOOCV) or other validation methods to assess the robustness of the selected variable model.

[0157] 3) Interpretation and application of results: Interpret the dimensionality reduction results and determine which underlying surface characteristic indicators are relatively independent and can be used for further analysis or model construction.

[0158] In practical applications, the results of different dimensionality reduction methods can be comprehensively considered in combination with professional knowledge and research purposes. In addition, the choice of dimensionality reduction method should be based on the characteristics of the data and the analysis objectives. For example, if the category distinction in the data set is very important, LDA may be a better choice; if the intrinsic structure of the data is of concern, LLE may be more appropriate; PCA is usually suitable for exploratory data analysis to identify the main components in the data.

[0159] According to the provided indicator system table (such as Table 1), the indicators in the underlying surface multi-dimensional characterization indicator system are divided into primary indicators, secondary indicators and tertiary indicators. The primary indicators refer to attribute characteristics, geometric properties and spatial distribution. Each primary indicator has its corresponding multiple secondary indicators, and each secondary indicator also has corresponding multiple tertiary indicators.

[0160] Furthermore, the first-level indicators and the corresponding second-level indicators and their third-level indicators are integrated, and the third-level indicators are weighted and linearly expressed in sequence to express the attribute characteristics, geometric characteristics and spatial distribution of the underlying surface. Assuming that each third-level indicator has been given a weight through data standardization, PCA, expert scoring and other methods, it can reasonably reflect the importance of each indicator in the comprehensive evaluation.

[0161] 1. Secondary indicators

[0162] For each secondary indicator under the primary indicator, we can calculate it according to the following formula:

[0163]

[0164] in, represents the comprehensive score of the secondary indicators under the primary indicator k, n k is the number of third-level indicators under the first-level indicator k, w ki is the weight of the third-level indicator i in the first-level indicator k, N ki is the standardized value of the third-level indicator i.

[0165] 2. Primary indicators

[0166] (1) Attribute characteristics of underlying surface (ATtribute):

[0167]

[0168] Among them, A T Represents the overall score of the underlying surface attribute characteristics, n T is the number of secondary indicators under the attribute characteristics, w Ti is the weight of the secondary index i in the attribute feature, s Ti is the standardized value of the secondary indicator i.

[0169] (2) Geometric characteristics of the underlying surface:

[0170]

[0171] Among them, A G Represents the overall score of the underlying surface geometric characteristics, n G is the number of secondary indices under geometric characteristics, w Gi is the weight of the secondary index j in the geometric characteristics, s Gi is the standardized value of the secondary indicator j.

[0172] (3) Spatial distribution of underlying surface (Dlstribution):

[0173]

[0174] Among them, A D Represents the overall score of the spatial distribution of the underlying surface, n D is the number of secondary indicators under spatial distribution, w Dk is the weight of the secondary index k in the spatial distribution, s Dk is the standardized value of the secondary indicator k.

[0175] 3. Comprehensive underlying surface index

[0176] Finally, we can combine the scores of the three first-level indicators to form a composite index for subsurface (CIS):

[0177] CIS=ω A ×AT+ω G ×GE+ω D ×DI

[0178] Among them, CIS represents the comprehensive index of the underlying surface of the basin in the study area, which comprehensively reflects the characteristics of the attribute characteristics, geometric characteristics, and spatial distribution in the basin; AT, GE, and DI are the values ​​of the three first-level indicators of underlying surface attribute characteristics, geometric characteristics, and spatial distribution, respectively. A ,ω G and ω D They are the weights of underlying surface attribute characteristics, geometric properties and spatial distribution.

[0179]

[0180]

[0181]

[0182] Although the above embodiments have been shown and described, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. Changes, modifications, substitutions and variations of the above embodiments by those of ordinary skill in the art are all within the scope of protection of the present invention.

Claims

1. A method for constructing a multi-dimensional indicator system for urban underlying surfaces, characterized in that: The following steps are involved: S1. Collect data, build an underlying surface database with high temporal and spatial resolution, and obtain annual LUCC data during the study period; S2. Based on the long-term annual LUCC data, the sub-basin grid is divided and the threshold of patch area is determined. The spatiotemporal evolution characteristics of the underlying surface within a given grid are analyzed and the characterization index is constructed. S3. Study the spatiotemporal evolution characteristics of multiple dimensional indicators and build a candidate library of underlying surface characteristic indicators; S4. Construct a multi-dimensional characterization index system for underlying surface; S5. Use data dimensionality reduction technology to process the data in the underlying surface multi-dimensional characterization index system and construct the underlying surface characterization index.

2. The method for constructing a multi-dimensional indicator system for urban underlying surfaces according to claim 1 is characterized in that: In the S1, the annual LUCC data within the study period is obtained. The specific process is to fuse the multi-source remote sensing image data of the complex underlying surface of the watershed including the urban area; extract and calculate the land use coverage / type (LUCC) data using the fused high-precision long-time series remote sensing images; Then, a natural-social environment auxiliary data set is established to study the dynamic development matrix of the underlying surface spatial pattern, and finally obtain the annual LUCC data during the study period.

3. The method for constructing a multi-dimensional indicator system for urban underlying surfaces according to claim 1, characterized in that: The multiple dimensional indicators in S3 include attribute features, geometric features and spatial distribution.

4. The method for constructing a multi-dimensional indicator system for urban underlying surfaces according to claim 3 is characterized in that: The steps of constructing the underlying surface multi-dimensional characterization index system in S4 are: determining the key indicators of each dimensional indicator, assigning appropriate weights to these key indicators, and developing a comprehensive scoring system.

5. The method for constructing a multi-dimensional indicator system for urban underlying surfaces according to claim 4 is characterized in that: The key characteristic indicators of each dimensional indicator in S4 are: attribute characteristics include terrain attributes, soil attributes, vegetation attributes, physical attributes and surface cover attributes; geometric characteristics include the area of ​​various land use types, the number of patches, the perimeter of patches and the complexity of patch shapes; spatial distribution includes connectivity, dispersion, aggregation and diversity.

6. The method for constructing a multi-dimensional indicator system for urban underlying surfaces according to claim 1, characterized in that: The data in the multi-dimensional characterization index system of the underlying surface is processed by using data dimensionality reduction technology in S5. The purpose is to identify the most representative and independent key indicators from a large number of original indicators through the evaluation of correlation and independence, so as to provide a solid data foundation for the final constructed characterization index.

7. The method for constructing a multi-dimensional indicator system for urban underlying surfaces according to claim 6 is characterized in that: The step of S5 comprises: S501. Data collection and preprocessing; S502. Data standardization; S503. Correlation analysis; S504. Data dimensionality reduction processing; S505. Dimensionality reduction screening and analysis of feature indicators.

8. The method for constructing a multi-dimensional indicator system for urban underlying surfaces according to claim 6 is characterized in that: The indicators in the underlying surface multi-dimensional characterization index system are divided into primary indicators, secondary indicators and tertiary indicators. The primary indicators refer to attribute characteristics, geometric properties and spatial distribution. Each primary indicator has corresponding multiple secondary indicators, and each secondary indicator also has corresponding multiple tertiary indicators.

9. The method for constructing a multi-dimensional indicator system for urban underlying surfaces according to claim 8, characterized in that: The underlying surface characterization index includes the primary index, the secondary index and the underlying surface comprehensive index. The three first-level indexes are expressed as follows: Underlying surface properties: Among them, A T Represents the overall score of the underlying surface attribute characteristics, n T is the number of secondary indicators under the attribute characteristics, w Ti is the weight of the secondary index i in the attribute feature, S Ti is the standardized value of the secondary indicator i; Geometric characteristics of underlying surface: Among them, A G Represents the overall score of the underlying surface geometric characteristics, n G is the number of secondary indices under geometric characteristics, w Gj is the weight of the secondary index j in the geometric characteristics, S Gj is the standardized value of the secondary indicator j; Spatial distribution of underlying surface: Among them, A D Represents the overall score of the underlying surface geometric characteristics, n D is the number of secondary indicators under spatial distribution, w Dk is the weight of the secondary index k in the spatial distribution, S Dk is the standardized value of the secondary indicator k.

10. The method for constructing a multi-dimensional indicator system for urban underlying surfaces according to claim 9, characterized in that: The scores of the three first-level indicators are integrated to obtain a comprehensive underlying surface index: CIS=ω A ×AT+ω G ×GE+ω D ×DI Among them, CIS represents the comprehensive index of the underlying surface of the basin in the study area, which comprehensively reflects the characteristics of the attribute characteristics, geometric characteristics, and spatial distribution in the basin; AT, GE, and DI are the values ​​of the three first-level indicators of underlying surface attribute characteristics, geometric characteristics, and spatial distribution, respectively. A ,ω G and ω D They are the weights of underlying surface attribute characteristics, geometric properties and spatial distribution.

Images