<p>Effective land utilization is a crucial strategy for mitigating carbon emissions in response to global climate change. Landscape pattern is intricately linked to land use carbon emissions (LCE). This study focuses on the urban agglomeration in the middle reaches of the Yangtze River (UAMRYR), a rapidly urbanizing yet ecologically sensitive region in central China. To elucidate how landscape pattern impact LCE, this study employed the eXtreme Gradient Boosting (XGBoost) model to evaluate the complex nonlinear influence mechanisms. The results revealed that from 2010 to 2020, the landscape shape, variety, and separation indices all increased, indicating an increase in landscape fragmentation. The total LCE from the UAMRYR have increased significantly, driven primarily by emissions from construction land. Spatial analysis revealed a "high-central, low-peripheral" emission pattern, with high LCE levels concentrated in the provincial capitals of Wuhan, Changsha and Nanchang. Increased landscape fragmentation and shape complexity—reflected in the SPLIT, DIVISION, LSI, AI, and SHDI values—are the primary drivers of LCE. Spatial heterogeneity exists across cities: in highly urbanized cities such as Wuhan, landscape fragmentation promotes carbon emissions, whereas in less developed cities such as Yiyang, moderate fragmentation and ecological diversification reduce LCE. Based on these findings, we derive spatial planning policy implications for cities at different development levels from a land use and landscape pattern perspective. This study aims to offer a methodological approach for regionally differentiated, landscape-informed carbon reduction planning.</p>