Rivers are globally significant sources of atmospheric carbon dioxide (CO2). However, the processes governing supersaturation of CO2 in large tropical fluvial networks are poorly understood. In particular, strikingly little is known about the role of land use in shaping CO2 variability in South Asian river basins, which are undergoing rapid urbanization. Here, we show that the wide variability in the partial pressure of CO2 (pCO2: 246.3-21271.2 mu atm) in an agriculture-dominated river basin (Krishna River basin, India) is primarily shaped by the extent of urbanization. Specifically, a strong positive correlation between pCO2 and built-up area (%) was observed when the built-up area exceeded 2%. Furthermore, machine learning analysis showed that pCO2 variability was predicted by built-up area (%), Strahler order, and altitude, together explaining similar to 77% of the spatial variability. Additionally, a strong negative correlation between excess CO2 and oxygen relative to atmospheric equilibrium indicates that in-stream metabolism, fueled by organic matter inputs from urbanized areas, is the primary cause of CO2 supersaturation, establishing a mechanistic link between pCO2 and the built-up area. Given that pCO2 increases with urbanization, limiting urban inputs is crucial for reducing fluvial CO2 emissions from South Asian river basins.