Regional Divergence in the Urban Form-Carbon Emission Nexus: A Comparative Analysis of Linear and Non-linear Spatial Modeling Approaches for 286 Chinese Cities

Sanqing He¹Yanan Sun^2*Ningyi Zeng³Lei Wang²Zihan Cao³Zhen He⁴

• ¹School of Design, Huazhong University of Science and Technology, Wuhan, China
• ²Architectural Design Institute, China Railway Construction Corporation Limited, Beijing, China
• ³School of Architecture and Urban Planning, Huazhong University of Science and Technology, Wuhan, China
• ⁴College of Literature and Journalism, Jishou University, Jishou, China

Urban areas account for over 70% of global CO₂ emissions. Recent studies show that urban morphology's impact on carbon emissions is more complex than traditional linear models suggest, with critical implications for low-carbon urban planning in China. This study contrasts linear spatial regression models with non-linear geographical Gaussian process regression (GGPR) to analyze morphology-emission relationships across 286 Chinese cities (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020). Linear spatial regression suggests that urban aggregation (UAI) consistently increases emissions with elasticity rising from 0.754 to 0.781, while synchronized vertical-horizontal development (UGSI) reduces emissions with tripling effectiveness from -0.096 to -0.297. However, GGPR with GeoShapley explainability exposes a fundamental paradox: the same morphological features show opposite effects depending on development stage and spatial context. While linear models and non-linear analysis both confirm UAI's complex role, they reveal it through different mechanisms. Linear models capture average effects across cities showing positive correlations, while GeoShapley analysis unveils location-specific variations where UAI can reduce per-capita emissions in certain contexts. Critical thresholds for key indicators like LPI remain relatively stable around 15, but cities' sensitivity to these thresholds has intensified dramatically. Eastern coastal megacities have reached saturation where further densification increases emissions, while western cities still benefit from compact development. The emergence of 3D indicators as significant factors by 2015-2020 marks a fundamental shift from 2D to 3D morphological influence. Based on the above findings, cities are suggested to implement threshold-based zoning that triggers mixed-use requirements at locally-calibrated limits and synchronized floor-area-ratio systems linking vertical development rights to infrastructure maturity. Also, performance-based carbon intensity targets need to replace uniform morphological standards, enabling regionally-differentiated strategies aligned with local development stages and spatial contexts.