Construction and Ecological Land Pattern Optimization for Enhancing Urban Ecosystem Services: A Focal Local Space Identification Perspective
Siwei Peng, Xingping WangOptimizing construction and ecological land patterns enhances urban ecosystem services, with efficiency gains depending on accurately identifying focal local spaces of greatest potential. However, existing research has yet to comprehensively quantify the multi-dimensional characteristics of both land types or systematically identify such priority areas. To address this gap, this study reconstructs integrated indicators capturing the “scale, shape, and connectivity” of construction and ecological land. Using the Local Climate Zone (LCZ) framework and interpretable machine learning, this study identifies focal local spaces for ecosystem service improvement in Nanjing, China, and analyzes the nonlinear interactive effects of these indicators on services. Key findings are as follows: (1) Construction land exhibits high aggregation and morphological complexity, whereas ecological land shows strong connectivity and intact morphology, with marked spatial differentiation. (2) Mid-rise building zones are identified as the focal local spaces (R2: 0.52–0.80). (3) Within these zones, ecological land core areas positively drive carbon storage (22.5%) and soil conservation (24.5%), construction land area proportion enhances water yield (23.7%) and construction land core areas suppress habitat quality (24.3%). Synergistic effects exist between ecological land extent and core areas for carbon and soil services, whereas construction land extent and core areas jointly degrade habitat quality. (4) This study recommends enhancing ecological land structure and connectivity, regulating construction land intensity, and creating semi-natural transitional zones at their interfaces. The integrated indicators and LCZ-based diagnostic approach provide a practical basis for zonal optimization of urban ecosystem services and a methodological reference for improving spatial governance.