A method for dividing grassland resource assets into homogeneous regions based on spatial continuous trees

Abstract

The present invention discloses a method for dividing grassland resource assets into homogeneous regions based on a spatial continuous tree. The present invention first selects the main indicators that affect the economic characteristics of grassland resources as attribute data for the division of homogeneous regions, thereby unifying the indicator system for the division of grassland resource assets into homogeneous regions; then, through spatial adjacency constraints and dynamically updated connection methods, combined with attribute differences, a spatial continuous tree is established to ensure that the final regionalization result is administratively complete, spatially continuous, and with similar attributes; finally, by setting the desired number of target regions, the spatial continuous tree is cut according to the principle of maximum heterogeneity reduction to form a regionalization result that meets the desired number. This method has no scale limit for the division of grassland resource asset homogeneous regions, and provides a reference for achieving higher-precision estimation of grassland resource asset value.

Description

Technical Field

[0001] The present invention relates to the field of establishing a grassland resource asset inventory price system, and in particular to a method for dividing grassland resource assets into homogeneous regions based on a spatial continuous tree. Background Art

[0002] The inventory of natural resource assets owned by the entire population aims to address prominent natural resource management issues, such as a lack of clear baseline data, and to explore ways to verify the rights and interests of state owners. This inventory includes both physical quantity statistics and valuation calculations. Establishing a pricing system for natural resource assets is a key and core task in this valuation process. The regionalization of homogeneous natural resource assets is crucial to the control and guidance standards for establishing a provincial pricing system and is a crucial step in establishing this system.

[0003] The "Technical Guidelines for the Inventory of Natural Resource Assets Owned by the Whole People" (Trial Implementation) are technical specifications developed for the inventory of natural resource assets owned by the whole people in my country. These guidelines define national homogeneous regions for state-owned agricultural land, forests owned by the whole people, grasslands, and marine resources in each province. However, the current delineation of homogeneous regions for natural resource assets lacks a unified indicator system and technical process. In particular, for grassland resources, national homogeneous regions are drawn solely based on existing industry zoning results. Furthermore, the delineation of national homogeneous regions in each province is relatively crude, with each province having only one to four national homogeneous regions, resulting in significant differences in resource status within some homogeneous regions. Finally, to meet the precision requirements of the inventory, the national homogeneous regions for each resource category are based on the county (district, city) as the smallest geographical unit (except for marine resources). This allows for the calculation of resource prices only at the county (district, city) level, resulting in a low level of refinement in the estimation of value.

[0004] The division of grassland resource assets into homogeneous regions is a new issue raised in the work of establishing a price system for the inventory of natural resources owned by the whole people. Existing technical methods use a combination of qualitative summary and quantitative investigation and analysis. There is no unified indicator system for the division of homogeneous regions. It cannot focus on the value of grassland resources for zoning and cannot well reflect the differentiation of grassland resource value. It is also impossible to obtain relatively detailed results of the division of grassland resource assets into homogeneous regions based on simultaneous consideration of attribute similarity and administrative boundary integrity.

[0005] Therefore, a regionalization method is needed that can unify the indicator system to reflect the differentiation of grassland resource values, consider attribute similarity and administrative boundary integrity, and reduce the scale of zoning to achieve the refinement and adjustment of the homogeneous areas of national grassland resource assets. Summary of the Invention

[0006] In order to solve the problems of the lack of a unified indicator system and technical process for the current division of grassland resource assets into homogeneous regions, the rough results of the division of national-level grassland resource assets into homogeneous regions, and the large differences in grassland resource conditions in the same homogeneous region, the present invention provides a method for the division of grassland resource assets into homogeneous regions based on attribute data such as grassland type, grassland area, hay yield, and grassland grade. This method can form regionalized results that meet the expected number, are spatially adjacent, and have similar attributes, and can realize the refinement of national-level homogeneous regions based on the actual regional conditions.

[0007] To achieve the above objectives, the present invention provides a method for dividing grassland resource assets into homogeneous regions based on a spatial continuous tree, comprising the following steps:

[0008] S1: Clarify the principles for selecting attribute data for the division of homogeneous regions of grassland resource assets, analyze the main factors affecting the economic characteristics of grassland resources, select grassland type, grassland area, hay yield and grassland grade as key indicators, and collect the basic data needed to calculate the relevant attribute data of the study area.

[0009] S2: Use Euclidean distance to measure the degree of difference between two geographical units. The degree of difference between two clusters is measured by the Euclidean distance between the two geographical units with the greatest difference (from two clusters respectively).

[0010] S3: Treat each geographical unit as a separate cluster, with spatial adjacency as the constraint, and calculate and update the Euclidean distance matrix according to the rules defined in step S2. Each update requires merging the two clusters with the smallest degree of dissimilarity. The update process is iterated until all geographical units are merged into one cluster.

[0011] S4: The cluster merging process is presented in the form of a dendrogram. Based on the cluster merging order of the dendrogram, a spatial continuous tree connecting all regional units is established. n regional units are connected by (n-1) edges.

[0012] S5: Use the sum of squared deviations (SSD R ) measures the heterogeneity of region R, takes heterogeneity reduction (ΔSSD) as the target optimization function for spatial continuous tree segmentation, and cuts the spatial continuous tree according to the principle of maximum heterogeneity reduction to obtain a regionalization result with k target regions.

[0013] As a preferred solution, step S1 specifies the following principles for selecting attribute data for homogeneous regional divisions of grassland resource assets: the attribute data must be measurable, universal, and easily accessible. To reflect the economic value of grassland resources in areas such as animal husbandry and forage production, grassland type, grassland area, hay yield, and grassland grade are selected as key indicators for both grassland land and grassland biodiversity. Grassland type and grassland area reflect changes in the physical volume of grassland land resource assets, while hay yield and grassland grade reflect grassland productivity and biological resource quality.

[0014] In step S1, the grassland resource attribute data monitoring is based on the specifications of NY / T 1233 "Technical Specifications for Grassland Resources and Ecological Monitoring", NY / T 1579 "Technical Specifications for Natural Grassland Grade Assessment" and NY / T 2997 "Grassland Classification".

[0015] Furthermore, in step S2, the Euclidean distance is used to measure the degree of difference between two geographical units. The degree of difference between two clusters is measured by the Euclidean distance between the two geographical units with the greatest difference (respectively from two clusters). The specific formula is as follows:

[0016]

[0017] d AB =max i∈A，j∈B (d ij )(2)

[0018] Where A and B are two clusters, i∈A, j∈B are two regional units; d ij is the Euclidean distance between regional units i and j; p is the number of attributes; x iv is the value of the vth attribute of regional unit i after Z standardization; x jv is the value of the vth attribute of regional unit j after Z standardization; d AB Indicates the degree of dissimilarity between clusters A and B.

[0019] Furthermore, in step S3, an n×n dimensional Euclidean distance matrix M (n represents the total number of regional units) is established based on the Euclidean distances between the regional units, and spatial adjacency constraints are used to focus on regional units with adjacent spatial positions.

[0020] In step S3, each geographical unit is treated as a separate cluster, and a pair of geographical units (assuming they are l and m) with the smallest Euclidean distance and spatially adjacent in the matrix M is found and merged to obtain cluster o. The degree of difference between other clusters and cluster o is calculated according to the dynamically updated connection method of formula (2) to update the matrix M; the above steps are repeated until all geographical units are merged into one cluster.

[0021] Furthermore, in step S4, each step of the cluster merging process is presented in the form of a dendrogram. Due to the spatial adjacency constraint, the clusters at any level of the dendrogram are spatially continuous; the dendrogram can reflect the degree of difference between clusters.

[0022] In step S4, a spatially continuous tree is established between the regional units according to the cluster merging order of the dendrogram; each time a regional unit is merged into a cluster, it is either connected to the only adjacent regional unit in the cluster or to the regional unit with the shortest Euclidean distance to it in the cluster; and finally (n-1) edges are formed to connect n regional units.

[0023] Furthermore, in step S5, the sum of squared deviations (SSD) R ) measures the heterogeneity of region R, and takes the overall heterogeneity reduction (ΔSSD) as the target optimization function for spatial continuous tree segmentation. The specific formula is:

[0024]

[0025] ΔSSD=SSD T -(SSD A +SSD B )(4)

[0026] In the formula, R represents a region, SSD R Indicates the size of its heterogeneity, p is the number of attributes, n r is the number of geographical units in region R, x ij It represents the value of the jth attribute of the i-th regional unit after Z-standardization. It represents the average value of the jth attribute of the regional unit in region R after Z standardization; ΔSSD represents the reduction of heterogeneity after splitting region T into regions A and B, and also represents the size of the variance between groups.

[0027] In step S5, the number of target regions k is set, and the spatial continuous tree is cut according to the principle of maximum heterogeneity reduction, that is, each cutting step makes the ΔSSD in formula (4) reach the maximum value, and the cutting is iterated (k-1) times to form k regions.

[0028] In step S5, in order to balance the area size of the final result, the minimum number of regional units contained in each area can be set as an additional constraint. If a certain cutting step can meet the principle of maximum heterogeneity reduction but cannot meet the additional constraint, this cutting step will not become a candidate cutting method, and the optimal cutting method that can preferentially meet the additional constraint will be selected.

[0029] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows:

[0030] The grassland resource asset homogeneous regional division method proposed in this invention has high application value, and its specific advantages are:

[0031] This method selects the main indicators that affect the economic characteristics of grassland resources as the attribute data for the division of homogeneous regions, thereby unifying the indicator system for the division of homogeneous regions of grassland resource assets. At the same time, this method can form a regionalized result that is spatially complete and continuous with similar attributes, meeting the requirements of complete administrative boundaries. In addition, this method is not limited to the scale for the division of homogeneous regions of grassland resource assets (the embodiment uses townships as the smallest administrative unit), which can achieve more refined grassland resource zoning and provide a reference for achieving more accurate estimation of the value of grassland resource assets. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is the township-level study area of the embodiment of the present invention;

[0033] Figure 2 A tree diagram formed by regionalization according to an embodiment of the present invention;

[0034] Figure 3 A spatially continuous tree formed by regionalization according to an embodiment of the present invention;

[0035] Figure 4 This is the regionalization result of the embodiment of the present invention;

[0036] Figure 5 It is a technical flow chart of the present invention. DETAILED DESCRIPTION

[0037] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0038] Example 1 [Taking grassland resource data in the western Sanjiangyuan area of Qinghai Province as an example]

[0039] The main factors affecting the economic characteristics of grassland resources include grassland type, grassland area, hay yield, and grassland grade. Based on the collected data, the township-level grassland area proportions (the region is mainly distributed with four types of grassland: alpine meadow, alpine steppe, alpine desert, and mountain meadow), the percentage of grassland area to township area, the average hay yield, and the average grassland grade score in the western Sanjiangyuan region of Qinghai Province (two counties and one town) were calculated. In other words, the seven attribute data values for each township were obtained and Z-standardized.

[0040] Each township is treated as a separate cluster, and the Euclidean distance between each township is calculated according to formula (1) to form a Euclidean distance matrix, as shown in Table 1. Considering the spatial adjacency constraint, spatially adjacent townships are marked in bold; among them, townships 1 and 3 have the smallest Euclidean distance (marked with *), and they are merged into a new cluster 1-3. The Euclidean distance matrix is updated according to formula (2), that is, d 1-3，i =max(d 1i , d 3i ),

[0041] where d 1-3，i is the degree of difference between the merged clusters 1-3 and township i, that is, the updated value in the Euclidean distance matrix, d 1i d 3i They represent the Euclidean distance between township 1 or 3 and township i before the merger, and the updated symmetric dissimilarity matrix is shown in Table 2.

[0042] Table 1 Initial Euclidean distance matrix

[0043] 1 2 3 4 5 6 7 8 9 10 11 12 13 1 0.00 1.87 0.75* 2.81 1.81 3.33 1.90 4.43 1.06 4.04 3.13 4.92 6.00 2 0.00 1.78 2.15 2.43 1.99 2.13 3.90 1.79 3.38 2.73 4.26 5.51 3 0.00 2.87 1.66 3.35 1.89 4.50 1.02 3.73 2.87 4.64 6.02 4 0.00 2.27 1.52 3.49 4.29 3.38 4.87 4.36 5.65 6.02 5 0.00 3.37 3.15 4.68 2.58 4.65 4.06 5.58 6.31 6 0.00 3.63 4.63 3.55 4.76 4.26 5.46 6.15 7 0.00 3.33 1.24 2.59 1.50 3.30 4.99 8 0.00 4.24 3.19 3.25 3.44 3.63 9 0.00 3.26 2.25 4.11 5.72 10 0.00 1.33 1.01 3.93 11 0.00 2.03 4.72 12 0.00 3.73 13 0.00

[0044] Table 2 Euclidean distance matrix after the first update

[0045] 1-3 2 4 5 6 7 8 9 10 11 12 13 1-3 0.00 1.87 2.87 1.81 3.35 1.90 4.50 1.06 4.04 3.13 4.92 6.00 2 0.00 2.15 2.43 1.99 2.13 3.90 1.79 3.38 2.73 4.26 5.51 4 0.00 2.27 1.52 3.49 4.29 3.38 4.87 4.36 5.65 6.02 5 0.00 3.37 3.15 4.68 2.58 4.65 4.06 5.58 6.31 6 0.00 3.63 4.63 3.55 4.76 4.26 5.46 6.15 7 0.00 3.33 1.24 2.59 1.50 3.30 4.99 8 0.00 4.24 3.19 3.25 3.44 3.63 9 0.00 3.26 2.25 4.11 5.72 10 0.00 1.33 1.01* 3.93 11 0.00 2.03 4.72 12 0.00 3.73 13 0.00

[0046] Continue to update the Euclidean distance matrix according to the above rules until all towns are merged into one cluster, and use a dendrogram to show the merging process, such as Figure 2 The length of the tree on the horizontal axis represents the Euclidean distance between regional units (or clusters).

[0047] A spatially continuous tree is built on the administrative region boundary map according to the cluster merging order of the dendrogram, such as Figure 3 The edge numbers represent the order in which the edges were created. When a town is merged into a cluster, it is either connected to the only neighboring town in the cluster or to the town with the shortest Euclidean distance to it. Towns are represented by serial numbers, with "-" used to indicate connecting two towns. The order in which the spatial continuous tree is generated is: 1-3, 10-11, 3-9, 4-6, 2-3, 10-12, 3-5, 7-10, 2-6, 8-13, 12-13, 9-11.

[0048] After establishing the spatial continuous tree, the edges of the tree need to be cut to form different regions. Considering the regional range and the number of towns, the target number of regions k = A is set, that is, the spatial continuous tree needs to be cut three times.

[0049] According to formula (3), the total SSD of the cluster is calculated (using SSD T denoted) and the SSD of the two clusters formed by cutting each edge (denoted by SSDA and SSD B denoted by ), and the heterogeneity reduction ΔSSD of the corresponding cutting is calculated according to formula (4). The calculation results are shown in Table 3. Find the cutting method that maximizes ΔSSD (in bold), that is, the cutting edge

[0050] Table 3 The first step of the cutting method selection process

[0051]

[0052] Cutting edge After that, two regions are formed and SSD is calculated in each of the two regions. T And cut each edge to form SSD A and SSD B , as shown in Table 4, where a blank row is used to separate the two regions. Find the cutting method (marked in bold) that maximizes ΔSSD, i.e., the cutting edge

[0053] Table 4 The selection process of cutting method in the second step

[0054]

[0055] Cutting edge After that, a total of 3 regions are formed, and SSD is calculated in each of the 3 regions. T And cut each edge to form SSD A and SSD B , as shown in Table 5, where a blank row is used to separate the two regions. Find the cutting method (marked in bold) that maximizes ΔSSD, i.e., cutting edge ⑨.

[0056] Table 5 The selection process of cutting method in the third step

[0057]

[0058] Cut edges one by one and ⑨, four grassland homogeneous regions with townships as the smallest administrative units were formed in the western part of Sanjiangyuan. The regionalization results are as follows Figure 4 .

[0059] According to the Technical Guidelines for the Inventory of Natural Resources Owned by the Whole People (Trial Version), the vast southern region of Qinghai Province, including the study area of this example, is included in a national homogeneous region (Qinghai Sanjiangyuan Grassland Area) (e.g. Figure 1). However, the region is vast, with large county areas and a transition zone between alpine grassland and alpine meadow. There are significant differences in grassland resource conditions between counties within the national homogeneous area and within the county itself. Since the construction of the grassland resource asset inventory price system is based on the working principle of "top-down and step-by-step adjustment", the construction of the grassland resource asset price system based on this result will inevitably lead to the grassland resource price levels in the counties within the homogeneous area seriously deviating from the previous level price (i.e., the national homogeneous area price), and will inevitably lead to large price differences between counties, which is contrary to the definition of a "homogeneous area". At the same time, subsequent work based on this result cannot distinguish the grassland prices of various townships within the county, and cannot reflect the differences in grassland resource asset values between townships.

[0060] The above-mentioned method of homogeneous regional division of grassland resource assets based on spatial continuous tree selects grassland type, grassland area, hay yield and grassland grade as key indicators for homogeneous regional division; determines the zoning scale as township level; establishes a spatial continuous tree through spatial adjacency constraints and dynamically updated connection methods, combined with attribute differences, to ensure that the final regionalization result is administratively complete, spatially continuous and with similar attributes; then considers the scope of the study area and the number of regional units, sets the target number of regions k=4, and cuts the spatial continuous tree according to the principle of maximum heterogeneity reduction to form a regionalization result that meets the expected number. Figure 4 The disclosed embodiment uses this method to divide the region into four homogeneous areas, which can scientifically and rationally realize the refinement of the national-level homogeneous areas. Compared with the national-level grassland homogeneous area zoning method of "based on existing zoning results, combining qualitative summary and quantitative investigation", this method can unify the indicator system for homogeneous area division, and realize high-efficiency grassland resource asset homogeneous area division with low data cost; and the national-level grassland homogeneous area uses district (county, city) as the smallest administrative unit for homogeneous area division, while this method has no limit on the scale for the division of grassland resource asset homogeneous areas (the embodiment uses township as the smallest administrative unit), providing a reference for achieving more accurate grassland resource asset value estimation.

Claims

1. A method for dividing grassland resource assets into homogeneous regions based on spatial continuous trees, characterized by: The following steps are involved: S1: Clarify the principles for selecting attribute data for the homogeneous regional division of grassland resource assets, analyze the main factors affecting the economic characteristics of grassland resources, select grassland type, grassland area, hay yield and grassland grade as key indicators, and collect the basic data needed to calculate the relevant attribute data of the study area; S2: Use Euclidean distance to measure the degree of difference between two geographical units. The degree of difference between two clusters is measured by the Euclidean distance between the two geographical units with the greatest difference from each other. S3: Treat each geographical unit as a separate cluster, with spatial adjacency as the constraint, and calculate and update the Euclidean distance matrix according to the rules defined in step S2. Each update requires merging the two clusters with the smallest degree of dissimilarity. The update process is iterated until all geographical units are merged into one cluster. S4: present the cluster merging process in the form of a dendrogram. Based on the cluster merging order of the dendrogram, a spatial continuous tree connecting all regional units is established. n regional units are connected by n-1 edges. S5: Use the sum of squared deviations (SSD) R Measure the heterogeneity of region R, use the heterogeneity reduction ΔSSD as the target optimization function for spatial continuous tree segmentation, cut the spatial continuous tree according to the principle of maximum heterogeneity reduction, and obtain a regionalization result with k target regions; In step S5, the sum of squared deviations SSD is used R The heterogeneity of region R is measured, and the overall heterogeneity reduction ΔSSD is used as the target optimization function for spatial continuous tree segmentation. The specific formula is: ΔSSD=SSD T -(SSD A +SSD B ) (4) In the formula, R represents a region, SSD R Indicates the size of its heterogeneity, p is the number of attributes, nr is the number of regional units in region R, x ij It represents the value of the jth attribute of the i-th regional unit after Z-standardization. It represents the average value of the jth attribute of the regional unit in region R after Z standardization; ΔSSD represents the reduction in heterogeneity after splitting region T into regions A and B, and also represents the magnitude of the between-group variance; In step S5, the number of target regions k is set, and the spatial continuous tree is cut according to the principle of maximum heterogeneity reduction, that is, each cutting step makes the ΔSSD in formula (4) reach the maximum value, and the cutting is iterated k-1 times to form k regions; In step S5, in order to balance the area size of the final result, the minimum number of regional units contained in each area is set as an additional constraint. If a certain cutting step can meet the principle of maximum heterogeneity reduction but cannot meet the additional constraint, this cutting step will not become a candidate cutting method, and the optimal cutting method that can preferentially meet the additional constraint will be selected.

2. The method for dividing grassland resource assets into homogeneous regions based on spatial continuous trees according to claim 1 is characterized by: In step S1, the principle of selecting attribute data for grassland resource asset homogeneous regional division is clarified: Attribute data must be measurable, universal, and easily accessible. To reflect the economic value of grassland resources in animal husbandry and forage production, grassland type, grassland area, hay yield, and grassland grade are selected as key indicators from the perspectives of grassland land and grassland biology, respectively. Grassland type and grassland area can reflect changes in the physical volume of grassland land resource assets; hay yield and grassland grade can reflect grassland productivity and biological resource quality. In step S1, the grassland resource attribute data monitoring is based on the specifications of NY / T 1233 "Technical Specifications for Grassland Resources and Ecological Monitoring", NY / T 1579 "Technical Specifications for Natural Grassland Grade Assessment" and NY / T 2997 "Grassland Classification".

3. The method for dividing grassland resource assets into homogeneous regions based on spatial continuous trees according to claim 1 or 2, characterized in that: In step S2, the Euclidean distance is used to measure the degree of difference between two geographical units. The degree of difference between two clusters is measured by the Euclidean distance between the two geographical units with the greatest difference, i.e., the two clusters. The specific formula is as follows: d AB =max i∈A，j∈B (d ij ) (2) Where A and B are two clusters, I∈A, j∈B are two regional units; d ij is the Euclidean distance between regional units i and j; p is the number of attributes; x iv is the value of the vth attribute of regional unit I after Z standardization; x jv is the value of the vth attribute of regional unit J after Z standardization; d AB Indicates the degree of dissimilarity between clusters A and B.

4. The method for dividing grassland resource assets into homogeneous regions based on spatial continuous trees according to claim 3 is characterized by: In step S3, an n×n dimensional Euclidean distance matrix M is established based on the Euclidean distances between the regional units, where n represents the total number of regional units, and spatial adjacency constraints are used to focus on spatially adjacent regional units. In step S3, each regional unit is treated as a separate cluster, and a pair of regional units with the smallest Euclidean distance and spatially adjacent in the matrix M is found, assuming they are l and m, and they are merged to obtain cluster o. The degree of difference between other clusters and cluster o is calculated according to the dynamically updated connection method of formula (2) to update the matrix M; Repeat the above steps until all geographical units are merged into one cluster.

5. The method for dividing grassland resource assets into homogeneous regions based on spatial continuous trees according to claim 1, 2 or 4, characterized in that: In step S4, each step of cluster merging is presented in the form of a dendrogram. Due to spatial adjacency constraints, clusters at any level of the dendrogram are spatially continuous; the dendrogram can reflect the degree of difference between clusters. In the step S4, a spatially continuous tree is established between the regional units according to the cluster merging order of the dendrogram; Each time a regional unit is merged into a cluster, it is connected to the only adjacent regional unit in the cluster, or connected to the regional unit with the shortest Euclidean distance to it in the cluster; eventually n-1 edges are formed to connect n regional units.

Images