Ten-Meter Resolution Spatial Allocation of CO2 Emissions in Tourist City Guilin using GOSAT Observations and Key Factor Attribution

This study takes Guilin, a typical tourist city, to explore its carbon emission characteristics. It integrates multi-source remote sensing data (satellite CO 2 column concentration, NO 2 , nightlight data, etc.), social statistics, geographic information of the road network, and constructs a 10-meter high-resolution spatial allocation product of CO 2 emissions for Guilin (2019-2024). The Logarithmic Mean Divisia Index (LMDI) is then adopted to quantify tourism-related driving factors. Under the most stringent buffered block cross-validation, the CatBoost model achieved an RMSE of 0.0010 kg CO 2 m -2 s -1 , an MAE of 0.0006 kg CO 2 m -2 s -1 , and an R 2 of 0.8225, while showing comparatively stable error performance across the spatially explicit validation schemes. Spatially, elevated predicted CO 2 emission flux density was concentrated around major tourism and transportation activity areas, including Liangjiang International Airport and the Lijiang River tourism corridor. Relative to ODIAC, the 10 m spatial allocation relocated the principal airport-related hotspot toward the actual airport area. COVID-19 mobility restrictions coincided with a decrease in mean CO 2 emission flux density of 2.31 × 10 -5 kg CO 2 m -2 s -1 in 2019–2020. The LMDI decomposition indicates that contraction of the tourism-scale effect was the dominant emission-reducing contribution during this interval. A modest decrease in mean CO 2 emission flux density was observed after 2023; however, the limited time series does not support attribution of this change to any specific policy intervention. The LMDI decomposition shows that industrial structure upgrading (cumulative contribution: −1.98 × 10 -4 kg CO 2 m -2 s -1 ) was the largest emission-reducing effect, whereas tourism-scale expansion (+8.60 × 10 -5 kg CO 2 m -2 s -1 ) partially offset this reduction. These contributions represent changes in the mean CO 2 emission flux density over the study area. The results highlight a “scale–structure” trade-off in tourism-related carbon emissions. This spatial allocation product can identify dispersed emission hotspots in tourism cities and provide methodological support for fine-scale spatial analysis. The results provide evidence for considering zone-specific emission management and differentiated low-carbon transition strategies in tourism-oriented cities, while the specific policy instruments require further feasibility assessment.