Quantifying nonlinear thresholds of grey-green space morphology on urban carbon metabolism across local climate zones: a machine learning study in Changchun city
Year: 2026
Type: article
Abstract: As core components of urban carbon metabolism, grey-green spaces, which refer to built-up and vegetated areas, exert a strong influence on atmospheric column-averaged CO₂ concentration (XCO₂) and net primary productivity (NPP), positioning them as a critical lever for low-carbon regulation. Yet, existing studies have largely concentrated on quantifying carbon sources and sinks at the city-wide scale, leaving the divergent drivers of carbon metabolism across different Local Climate Zones (LCZ) largely unexplored. To address this gap, we conducted a case study in Changchun, China. Drawing on the LCZ classification framework, we identified LCZ 1 (compact build-up area) and LCZ 6 (open low-rise area) as our focal zones, as these two classes yielded the highest predictive accuracy for both XCO₂ and NPP. We then integrated multi-source geospatial data to construct a multidimensional morphological indicator system tailored to grey-green spaces. Using CatBoost-SHAP—an interpretable machine learning approach—we developed separate predictive models for XCO₂ and NPP within these two LCZs, allowing us to uncover the nonlinear mechanisms through which urban morphology shapes carbon metabolism. Our results indicate that: (1) In LCZ 1, XCO₂ and NPP are predominantly driven by grey-space morphological factors (B_CORE, BD, BH), with a combined contribution exceeding 55%; in LCZ 6, by contrast, green-space factors (VFC, ACH) together with climatic background (MAT) dominate, accounting for 66.8% and 42.3% of the contributions, respectively. (2) Key factors exhibited significant nonlinear thresholds. In LCZ 1, optimal NPP was observed when B_CORE < 2.0 and BH ranged between 10 and 40 m. In LCZ 6, VFC > 0.65 exacerbated XCO₂, while B1 showed a U-shaped inflection point at -0.25 in its effect on NPP. (3) The same factor exhibited functional differentiation. VFC served as a proxy for human activity intensity in LCZ 6, while acting as an indicator of vegetation carbon sink in LCZ 1. B_CORE played a dual role in LCZ 1, suppressing XCO₂ while enhancing NPP.This study developed a technical framework integrating LCZ classification, spatial pattern quantification, mechanism analysis, and zoning planning. The framework enabled us to pinpoint key regulatory thresholds, providing quantitative evidence and actionable planning pathways to support the low-carbon transition of cities in the future.
Source: Scientific Reports
Authors: Yuqing Gong, Wenjun Qin
Institution: Shenyang Jianzhu University
Language: English
Topic: Urban Heat Island Mitigation
Subfield: Environmental Engineering
Field: Environmental Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7223215874