CN-122020306-A - Ecological corridor identification method and system based on machine learning

CN122020306ACN 122020306 ACN122020306 ACN 122020306ACN-122020306-A

Abstract

The invention relates to an ecological corridor identification method and system based on machine learning. The method comprises the steps of firstly extracting an ecological source land based on land utilization data, training a machine learning classification model by utilizing ecological system service data and multi-source space influence factors, inputting the multi-source space influence factors into the trained machine learning classification model, predicting ecological resistance values of all space units in a research area, outputting an ecological resistance surface, inputting the ecological source land and the ecological resistance surface into an ecological connectivity analysis model based on a circuit theory, simulating an ecological flow process, identifying space channels connected between different ecological source lands according to ecological flow probability or accumulated resistance, and determining an ecological corridor in the research area. Compared with the prior art, the method has the advantages of improving the objectivity and the accuracy of the ecological corridor identification result and the like.

Inventors

QIU JIANG
CHEN TIANYU

Assignees

上海应用技术大学

Dates

Publication Date: 20260512
Application Date: 20260130

Claims (10)

1. The ecological corridor identification method based on machine learning is characterized by comprising the following steps: S1, land utilization data of a research area are obtained, and ecological source land is extracted based on the land utilization data; S2, acquiring ecosystem service data and multisource space influence factors of a research area, training a machine learning classification model by using the ecosystem service data and the multisource space influence factors, inputting the multisource space influence factors into the trained machine learning classification model, predicting ecological resistance values of all space units in the research area, and outputting an ecological resistance surface; S3, inputting the ecological source land and the ecological resistance surface into an ecological connectivity analysis model based on a circuit theory, simulating an ecological flow process, identifying space channels connected between different ecological source lands according to ecological flow probability or accumulated resistance, and determining an ecological corridor in a research area.
2. The machine learning-based ecological corridor identification method according to claim 1, wherein the specific step of extracting the ecological source and land in S1 includes: reclassifying land utilization types of the research area into ecological land and non-ecological land, and extracting ecological land patches as candidate areas; Carrying out image morphology processing on the candidate region by using a morphology space pattern analysis method, and screening out plaque with morphology type as a core region as an ecological source candidate object; And carrying out connectivity evaluation or area threshold screening on the ecological source and ground candidate objects, and determining the plaque meeting the preset condition as the ecological source and ground.
3. The machine learning based ecological corridor identification method of claim 1, wherein the ecosystem service data in S2 includes water yield, carbon reserves, habitat quality, and soil conservation data.
4. A machine learning based ecological corridor identification method as claimed in claim 3, wherein the specific step of training the machine learning classification model in S2 includes: carrying out normalization processing on the ecosystem service data, and constructing comprehensive ecosystem service grade distribution by utilizing the normalized ecosystem service data; marking the area with the lowest grade in the comprehensive ecosystem service grade distribution as a negative sample sampling area, and marking the ecological source as a positive sample sampling area; And randomly selecting a plurality of sample points in the negative sample sampling area and the positive sample sampling area respectively, taking a multisource space influence factor corresponding to the selected sample points as a characteristic variable, taking the category attribute of the sample points as a tag variable, constructing a training data set, and training a machine learning classification model by using the training data set.
5. The machine learning based ecological corridor identification method of claim 4, wherein the machine learning classification model in S2 is a gradient lifting based model, and the machine learning classification model specifically comprises XGBoost, catBoost or LightGBM.
6. The machine learning based ecological corridor identification method of claim 4, wherein in the process of training the machine learning classification model, an automatic super-parameter optimization method is adopted to optimize model parameters.
7. The machine learning based ecological corridor identification method of claim 1, wherein the multi-source spatial impact factors in S2 include natural environment factors and human activity factors; the natural environment factors comprise one or more of elevation, gradient, topography fluctuation, normalized vegetation index and rainfall; The human activity factor includes one or more of a distance to road, a distance to water, population density, night light intensity, human footprint index, and a domestic total production value.
8. The machine learning-based ecological corridor identification method according to claim 1, wherein the specific process of predicting the ecological resistance value of each space unit in the research area and outputting the ecological resistance surface in S2 includes: Dividing a research area into a plurality of regular grid cells, generating a prediction point at the center of each grid cell, matching corresponding influence factor values for each prediction point, and constructing a prediction data set; Inputting the predicted data set into a trained machine learning classification model, and outputting a predicted ecological resistance value set; Carrying out space association processing on each ecological resistance value in the ecological resistance value set, associating each ecological resistance value with a corresponding regular grid unit, and generating spatial prediction result data; and converting the predicted result data into a grid data format to form an ecological resistance surface of the research area.
9. The machine learning-based ecological corridor identification method according to claim 1, wherein the circuit theory-based ecological connectivity analysis model in S3 is further used for identifying ecological pinch points, and the specific steps of identifying ecological pinch points are as follows: calculating the ecological flow density of each space unit in the research area based on the ecological flow process simulated by the ecological connectivity analysis model; Identifying a region with the ecological flow density higher than a preset threshold as an ecological pinch point; the ecological connectivity analysis model based on the circuit theory in the step S3 is also used for identifying ecological barrier points, and the specific steps of identifying the ecological barrier points are as follows: calculating ecological flow improvement coefficients of the space units in the research area based on the ecological connectivity analysis model, wherein the ecological flow improvement coefficients represent the improvement degree of ecological flow connectivity after removing the ecological resistance of the space units; and identifying the area with the ecological flow improvement coefficient higher than a preset threshold as the ecological barrier point.
10. An ecological corridor identification system based on machine learning, which is characterized in that the system works by applying the ecological corridor identification method based on machine learning according to any one of claims 1-9, and the system comprises an ecological source and ground identification module, an ecological resistance surface construction module and an ecological corridor identification module; the ecological source land identification module is used for extracting and outputting an ecological source land based on land utilization data of an acquired research area; The ecological resistance surface construction module is used for acquiring the ecological system service data and the multisource space influence factors, training a machine learning classification model based on the ecological system service data and the multisource space influence factors, predicting the ecological resistance value of each space unit by combining the trained model with the multisource space influence factors, and outputting an ecological resistance surface; The ecological corridor identification module is used for inputting the ecological source and the ecological resistance surface into an ecological connectivity analysis model based on a circuit theory, simulating an ecological flowing process and identifying an ecological corridor between the ecological source and the ecological resistance surface.

Description

Ecological corridor identification method and system based on machine learning Technical Field The invention relates to the technical field of ecological protection planning, in particular to an ecological corridor identification method and system based on machine learning. Background With the continuous acceleration of the global urbanization process, the natural ecological space is continuously extruded, the problem of habitat disruption is increasingly prominent, and the protection of biodiversity and the maintenance of the service functions of an ecological system face great challenges. Under the background, the dispersed and isolated ecological patches are effectively connected through scientific identification and construction of the ecological corridor, and the ecological patch has important significance in guaranteeing regional ecological safety patterns, promoting species migration and gene communication and improving the overall stability of an ecological system, and has become key technical content in the fields of homeland space planning and ecological protection and restoration. The existing ecological corridor identification method is generally based on a technical framework of ecological source and ground, ecological resistance surface and ecological corridor, wherein the construction of the ecological resistance surface is used for representing the obstruction degree of different space units to the ecological flow process, and is a core link for influencing the accuracy and reliability of corridor identification results. However, the existing resistance surface construction method has the defects that firstly, the resistance value and the weight are subjected to subjective assignment by depending on expert experience, the result has strong artificial property, consistency and repeatability of results among different researchers or different areas are difficult to ensure, secondly, a weight system set by fixing or experience is often adopted, differences between ecological environment conditions of different areas and human interference characteristics are difficult to effectively reflect, the application range is limited, and thirdly, the complex nonlinear relation shown by an ecological process is difficult to characterize based on linear superposition assumption. For example, the invention patent with publication number of CN119692602A discloses an ecological restoration quantization method based on a river basin ecological safety complex network, which selects five characteristic indexes of river basin topography conditions, surface coverage, artificial interference, vegetation state and ecological benefit to construct a river basin ecological resistance surface, and constructs an ecological safety complex network model on the basis. Although the method integrates multiple influencing factors, the method essentially belongs to a construction method based on a preset index system. In determining the resistance value for each spatial element, the method typically relies on ranking, empirical assignment, or simple superposition calculations of the five selected characteristic indices. The method lacks a training process based on real sample data, is difficult to objectively quantify the specific contribution degree of each index to ecological resistance, and is difficult to capture complex and nonlinear joint stress effects on ecological flow caused by multi-source environmental factors (such as interaction between artificial interference and vegetation states) in a superposition mode based on an index system, so that the generated ecological resistance surface still has room for improvement in objectivity and precision. In conclusion, the construction of the ecological resistance surface in the current ecological corridor identification has the objective quantification difficult problems that subjective assignment is carried out by severely relying on expert experience, and the nonlinear relation among multiple source factors cannot be effectively represented. Disclosure of Invention The invention aims to overcome the defects of the prior art and provide an ecological corridor identification method and system based on machine learning. The aim of the invention can be achieved by the following technical scheme: According to one aspect of the invention, there is provided an ecological corridor identification method based on machine learning, the method steps comprising: S1, land utilization data of a research area are obtained, and ecological source land is extracted based on the land utilization data; S2, acquiring ecosystem service data and multisource space influence factors of a research area, training a machine learning classification model by using the ecosystem service data and the multisource space influence factors, inputting the multisource space influence factors into the trained machine learning classification model, predicting ecological resistance values of all space units in th