Highlights

• The study introduces a novel SSA-BP neural network approach to estimate CO₂ emissions at the provincial level and downscale to the city level using an autoencoder.
• It creates a comprehensive, city-level CO₂ emissions inventory for GBA cities spanning two decades (2000-2019), addressing previous data gaps.
• The research provides insights into spatial and temporal dependencies in emissions data, revealing regional clusters and specific influential factors across time.

Summary

The research article investigates CO₂ emissions in the Guangdong-Hong Kong-Macao Greater Bay Area (GBA) and surrounding cities. It addresses the challenge of estimating city-level CO₂ emissions due to limited data by applying a neural network model, SSA-BP, optimized with a sparrow search algorithm, and an autoencoder to downscale provincial emissions estimates to city-level data.

The findings reveal a spatial pattern of emissions, with high emissions concentrated in central GBA cities such as Guangzhou and Shenzhen, and lower emissions in peripheral areas. The study also identifies factors influencing emissions, with GDP, population, and energy intensity exerting positive effects, while technological innovation contributes negatively.

These results provide insights for urban-level carbon reduction strategies, emphasizing the importance of economic restructuring, population distribution, and technological advancement in managing emissions​.

X. Luo, C. Liu, and H. Zhao, “Modeling and spatio-temporal analysis on CO2 emissions in the Guangdong-Hong Kong-Macao greater bay area and surrounding cities based on neural network and autoencoder,” Sustainable Cities and Society, vol. 103, p. 105254, Apr. 2024, doi: 10.1016/j.scs.2024.105254.