Construction and Optimization of Ecological Security Patterns Based on Som and Multi-Source Data at Different Scales: A Case Study of Dongting Lake Basin in China

Abstract

Global climate change and intensified human activities have exposed regions to complex ecological risks. Constructing ecological security patterns (ESPs) is crucial for ensuring regional ecological security and optimizing territorial space. However, traditional research methods homogenize ecological sources by failing to extract original information from factors, and most studies construct ESPs at a single scale while neglecting scale effects in ecological source identification. Therefore, this study selected the Dongting Lake Basin as the research area, introducing Self-Organizing Mapping Neural Network (SOM) model into an "importance-sensitivity-connectivity" framework at both grid and sub-watershed scales to identify important ecological sources. The ecological resistance surface was modified using socio-ecological factors, and circuit theory was employed to construct ESPs while formulating spatial optimization strategies. Key findings include: (1) Ecological sources showed similar spatial distributions at both scales, concentrating in western mountainous areas and basin peripheries, with notable absence in urbanized central regions. However, there were significant differences in the number, area, connectivity of sources at the two scales. A total of 38 ecological sources covering 45,035.47 km2 were determined at the grid scale. In contrast, the sub-watershed scale revealed 28 ecological sources with a total area of 92,349.85 km2, and the landscape connectivity index of the patches was greater. (2) High-resistance areas clustered in central urban zones and eastern plains of the basin, while low-resistance regions occurred in eastern and central areas with gentle slopes and high vegetation coverage. (3) Both scales exhibited sparse ecological corridors in central regions, with 90 corridors (3,779.51 km) predominantly in western areas at the grid scale, while 56 optimal corridors (2,256.59 km) concentrated in eastern areas at the sub-watershed scale. (4) 129 and 45 ecological nodes (pinch points and barriers) were extracted at the grid and sub-watershed scales, respectively. At the grid scale, they mainly distributed on the ecological corridors in the west and south of the study area, while they mainly distributed on the ecological corridors in the east and south at the sub-watershed scale. (5) Spatial optimization patterns were established as "one ring, two corridors, three zones, multiple cores" at the grid scale and "one ring, one corridor, three zones, multiple cores" at the sub-watershed scale. This required differentiated spatial optimization measures that emphasized a holistic approach to conserving mountains, rivers, forests, farmlands, lakes, and grasslands, aligned with regional characteristics and long-term planning. This study innovatively integrates ecological functional zoning and holistic element governance concepts, providing valuable references for regional land development, ecological restoration, and environmental management.

Keywords: Ecological security pattern, Self-organizing mapping neural network, Ecological source, Ecosystem service, Dongting Lake Basin