Spatiotemporal Association Pattern Mining and Evolution Simulation of Urban Carbon Emissions

Abstract

In response to the challenge of continuous carbon emission growth, China has set goals of achieving carbon peak and carbon neutrality. Urban carbon emission association analysis and evolution process simulation aim to understand the driving mechanisms of urban carbon emissions, providing a scientific basis for precisely formulating urban carbon reduction policies. The evolution process of urban carbon emissions is influenced both by spatial associations between cities, such as population mobility and industrial collaboration, as well as by temporal associations of multiple internal dimensions including population, economy, and built-up area expansion. However, existing studies mostly model the evolution process from the perspective of a single city or assume multiple cities have identical carbon emission driving mechanisms. These studies have three major deficiencies: (1) they ignore the spatial transfer effects of carbon emissions between distant cities, (2) they make it difficult to identify regions with closely associated carbon emission changes, and (3) they insufficiently recognize the heterogeneity of temporal driving factors of urban carbon emissions. These cognitive deficiencies in spatiotemporal association patterns directly affect the accuracy and reliability of urban carbon emission evolution process simulation. Therefore, this thesis focuses on the joint driving mechanisms of spatial association patterns and multi-dimensional temporal associations on urban carbon emissions, conducting research on spatiotemporal association pattern mining and evolution simulation at the city scale. The following are the main contents of this thesis.

(1) Addressing the limited understanding of association patterns between carbon emissions of distant cities, this thesis extracts carbon emission events (such as growth and decline) from each city’s time series. A spatial association pattern mining method is then constructed, coupling Euclidean distance, migration network distance, and semantic distance constraints. Experiments conducted in Chinese cities (excluding Hong Kong, Macao, and Taiwan) revealed the spatial association relationships between different types of urban carbon emission events. Compared with existing co-location pattern mining methods and ablation experiments, results demonstrate that the proposed method can identify more comprehensive spatial association patterns and reduce pattern omissions caused by hard segmentation of local potentially associated regions.

(2) Addressing the problem that existing studies have not fully recognized the differences in connection strengths between cities and their surrounding cities in urban carbon emission association networks, this thesis develops a hotspot and coldspot identification method. The method quantifies the association strength of each city in local regions based on community detection and network constraint coefficients. It then couples spatial neighborhood constraints with attribute similarity constraints to identify regions where urban carbon emissions show close associations. Experiments conducted in Chinese cities (excluding Hong Kong, Macao, and Taiwan) revealed a spatial pattern characterized by weak carbon emission association strength among cities in China’s eastern coastal regions. Furthermore, compared with existing hotspot and coldspot detection methods, results indicate that the proposed method improves the accuracy of identifying regions with high and low association strength.

(3) Addressing the problem that existing studies have not fully recognized the differences in carbon emission driving factors for various types of cities at different development stages, this thesis proposes a strategy for discovering driving factors of urban carbon emission temporal changes under spatiotemporal hierarchical constraints. First, city clusters are determined based on both the temporal growth trends of carbon emissions in each city and geographical environmental factors; then, within each cluster, Multispatial Convergent Cross Mapping techniques and meta-analysis methods are applied to test the temporal association relationships between various driving factors and urban carbon emissions. Through empirical analysis of 101 Chinese cities, the proposed method identified four city clusters: three clusters in the growth stage (industry-oriented, service-oriented, and intensive development cities) and one cluster in the fluctuating optimization stage (primary industry-oriented cities), while effectively uncovering the carbon emission driving factors for each cluster.

(4) Existing studies have insufficiently considered the impact of spatial association patterns between cities and the spatiotemporal differentiation of carbon emission driving factors on carbon emissions, leading to inaccurate simulation of urban carbon emission evolution processes. This thesis proposes a ‘regional-urban’ carbon emission evolution simulation method that couples spatiotemporal association knowledge. Differentiated simulation parameter setting mechanisms are constructed for each city and a regional-scale carbon emission target constraint mechanism is introduced to quantify the feedback of spatial associations between cities on the regional carbon emission system. Empirical studies in the Beijing-Tianjin-Hebei and Yangtze River Delta urban agglomerations show that the proposed method generates simulation strategies that can effectively reduce urban carbon emissions or advance carbon emission peak times.

This thesis deepens the understanding of spatiotemporal association patterns of urban carbon emissions and achieves scientific modeling of urban carbon emission evolution processes. It provides reliable methodological support and scientific decision-making basis for achieving China’s carbon peaking and carbon neutrality goals.

Date of Award	4 Sept 2025
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	W. Z. LU (Supervisor), Min Deng (External Supervisor), Siu Ming LO (Supervisor) & Xiaowei LUO (Co-supervisor)