Cities are typical sources of carbon and a focus of climate-change mitigation. Although there is great potential for reducing emissions in cities, constructing low-carbonemission cities under the carbon-emission reduction target of the 2016 Paris Agreement remains challenging as there is little scientific evidence for use by government decisionmakers. To solve this problem, we must understand the urban carbon-cycle process in greater detail and especially analyze spatiotemporal variations in carbon sources and sinks. Previous studies have focused on developing methods for estimating urban carbon sinks, such as remote sensing retrieval algorithms, a light-use efficiency model, and an ecological process model. In studies of urban carbon sources, night-light imagery has been widely used to construct regional-, national- and global-scale carbon emission maps. Additionally, open $CO_2$-emission products have provided some policy implications for decision-makers. Both the research and associated products have yielded fruitful results; however, the studies are independent and there are few that consider both the ecological process model and the $CO_2$-emission inventory. Therefore, we present a method that couples the ecological process model with an emission inventory to map the spatial distribution of carbon sources and sinks. We integrated multiple sources of data with different models and analytical methods to analyze the carbon balance status of Jinjiang city in 2013. The results indicated that a combined machine learning and spatial statistical approach may improve the accuracy of estimating the regional net primary productivity. Further, we upscaled the Biome-BGC model and assimilated the data to reduce the uncertainty of the upscaling process. Long-term observations in the data assimilation framework may improve the accuracy of the ecological process model’s simulation. Moreover, we may be able to devise a general hybrid approach (“globally downscaled” and ” bottom up” ) to create a gridded map showing $CO_2$ emissions. Comparisons between the resulting maps and other open products indicate a high accuracy in the maps. Finally, the results indicated that Jinjiang city was in a carbon-loss state in 2013 (that is, more carbon was emitted than stored). Carbon sources were distributed mainly in urban and suburban areas, whereas carbon sinks were distributed mainly in suburban areas. The population density, average temperature, and fragmentation of urban functional districts all had a significant effect on the urban carbon surplus. The population density negatively correlated with the carbon surplus. The average temperature and fragmentation of urban functional districts varied across the city. The results may inform policies on such issues as population migration, walkability, environmental management, construction, and urban renewal and thus help decision-makers mitigate carbon loss in cities.