Identification of Redundant Patches in Early Urbanized Areas Based on mRMR + SVM

Abstract

Rapid global urbanization has inevitably led to habitat fragmentation, and much research has focused on responding by building ecological networks and improving connectivity. The construction of such post-event ecological networks is sometimes not effective, and the identification and protection of important patches in the early urbanization period is more significant. Based on the redundancy theory, this paper explores the method of identifying potential critical habitat patches after future urban development. The paper takes the Longxing area of Chongqing city as an example and uses the cumulative current value results of circuit theory model simulations and patch characteristic attribute information as sample data, introduces a combination of the minimal-redundancy-maximal-relevance criterion and a support vector machine (mRMR + SVM) to discriminate the redundancy of patches in ecological networks, and ranks the importance of source patches. The results show that the mRMR criterion can exhibit a more realistic ranking of patch importance, allowing patches with better quality to achieve a higher ranking. Compared to the patch importance determined based on the magnitude of the cumulative current value, 66.34% of the patch importance order changed in the mRMR criterion ranking results. The mRMR + SVM method was effective in identifying individuals with redundancy from the set of source-site patches. The test found that ecological corridors still passed through or occurred near these redundant patches after their removal, demonstrating the substitutability of patches identified as redundant. The results of the study can help improve the knowledge of ecological networks and provide quantitative methodological support and decision-making reference for urban development and ecological conservation.