A Machine Learning Approach to Relate Green Space Landscape Metrics to Net Primary Production Across Shanghai’s Built Environment

Rongxiang Chen, Xunrui Ou (), Mingjing Xie, Zixi Chen and Kaida Chen ()
Additional contact information
Rongxiang Chen: College of Landscape Architecture and Art, Fujian Agriculture and Forestry University, Fuzhou 350002, China
Xunrui Ou: College of Landscape Architecture and Art, Fujian Agriculture and Forestry University, Fuzhou 350002, China
Mingjing Xie: College of Landscape Architecture and Art, Fujian Agriculture and Forestry University, Fuzhou 350002, China
Zixi Chen: College of Architecture, Tianjing University, Tianjin 300072, China
Kaida Chen: College of Landscape Architecture and Art, Fujian Agriculture and Forestry University, Fuzhou 350002, China

Land, 2025, vol. 14, issue 12, 1-34

Abstract: Achieving carbon neutrality has become one of the core objectives in contemporary urban development and sustainable growth, underscoring the importance of clarifying the relationship between urban green space landscape metrics and plant carbon sequestration. While existing research confirms the significant role of the structure and pattern of green spaces in carbon sequestration, systematic understanding of their relationship at the local scale within diverse built environments remains limited. To address this, this study objectively categorises five types of built environments using K-means clustering and conducts in-depth analysis on four representative areas. Employing the CatBoost machine learning model and the Shapley Additive Propensity (SHAP) method, we highlighted the influence of green space pattern characteristics on net prmary productivity (NPP) across different built environments. The findings are as follows: (1) Green Coverage Ratio (GCR) exhibits the highest contribution among all explanatory variables across different built environments. In low-intensity built environments, it contributes 74% to the overall explanation, showing a stable association between higher green space proportion and higher carbon sink levels. (2) In high-intensity built environments, limited green spaces exhibit a pronounced “spatial compensation effect” through morphological optimisation and enhanced spatial connectivity. In medium-intensity built environments, they demonstrate a “moderate positive effect,” with peak carbon sequestration efficiency occurring when GCR ranges from 0.25 to 0.75, aggregation index (AI) from 94 to 98, and splitting index (SI) from 1.2 to 1.4. (3) Significant interactions exist among green space landscape metrics, with moderately connected and moderately complex spatial structures enhancing carbon sink efficiency. This study reveals the differentiated impact by which green space landscape metrics influence carbon sink effects under varying urban built environments, providing scientific basis for optimising urban green space systems and low-carbon spatial planning.

Keywords: carbon sink; urban planning; machine learning; green space landscape metrics; shapley additive explanation; built environment; CASA model (search for similar items in EconPapers)
JEL-codes: Q15 Q2 Q24 Q28 Q5 R14 R52 (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
https://www.mdpi.com/2073-445X/14/12/2349/pdf (application/pdf)
https://www.mdpi.com/2073-445X/14/12/2349/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:gam:jlands:v:14:y:2025:i:12:p:2349-:d:1806886

Access Statistics for this article

Land is currently edited by Ms. Carol Ma

More articles in Land from MDPI
Bibliographic data for series maintained by MDPI Indexing Manager ().