Optimization of cold island network and thermal environment mitigation in urban agglomeration based on complex network theory in central Guizhou Province, southwestern China

Objective The thermal environment is a critical metric for assessing urban ecological quality, directly impacting residents^’ health and regional sustainable economic development. Focusing on the central Guizhou urban agglomeration in southwestern China, this study overcame the limitations of traditional isolated research on blue-green space morphology and innovatively proposed a cold island network optimization framework, aiming to maximize thermal risk mitigation with minimal intervention.

Method Based on MODIS LST data, the cold island network was extracted using morphological spatial pattern analysis (MSPA), minimum area threshold analysis, and the ant colony algorithm. From complex network theory, the shortcut for maximum betweenness (SMB) and reduction structural hole (RSH) strategies were introduced for edge-addition optimization, with structural robustness analysis employed to verify effectiveness. The importance levels of “source” patches were classified via weighted fusion of network structure parameters, and obstacle points were identified by overlaying heat island patches.

Result (1) Significant spatial heterogeneity in the agglomeration’s thermal environment: high-temperature zones were concentrated in the north-central and southeastern regions, while low-temperature zones were located in the south and southwestern region. (2) A total of 102 cold island “sources” (5 518.412 km²) were identified, dense in the south and sparse in the north, clustered continuously in the southern and southwestern region and scattered in the northern region. 190 corridors (total length 2 791.58 km) were extracted, with overall weak connectivity and loose structure. (3) The optimization effect of SMB was better than RSH: initial connection robustness increased from 0.73 to 0.82. Under deliberate attacks, the threshold number of node removals required for complete recovery of network nodes and edges rose from 9 and 13 to 22 and 23, respectively. Under random attacks, the threshold number of node removals increased from 11 and 12 to 23 and 25, respectively. Recovery robustness improved significantly, resulting in a ring-radial network pattern. (4) Six level-5 sources (46.88% of total source area, dominated by large water bodies and mountains) and 16 obstacle points (11 blocking existing corridors, 5 in new optimized ones) were identified.

Conclusion This study provides a scientific and efficient optimization model for enhancing thermal resilience in karst mountain cities, offering a new paradigm for precise ecological space regulation in sustainable high-density urban development.