Real-time characterization model of carbon emissions based on land-use status: A case study of Xi'an city, China

The traditional carbon accounting method, with a lag of over 2 years due to the release time of statistical yearbooks, impedes timely policy adjustments in urban planning and management. Hence, there is an urgent need to establish a real-time carbon emissions characterization model. Xi'an which has a complex land-use structure was chosen as the study site and its carbon emissions were calculated using the Emission Factor Method. The GIS-Kernel Density (KD) model was constructed, and land use was subdivided based on Point of Interest (POI) and road network data. Based on the results of carbon emissions accounting and land-use subdivision, a Multilayer perceptron (MLP) model was established. The RS images of Xi'an underwent supervised classification, and the carbon emissions of Xi'an were characterized based on the subdivision results and MLP model. The results show that: (1) The accuracy of the characterization model is more than 90%, and with the improvement of remote sensing technology, the accuracy will be further improved; (2) Compared with the existing model, this model can real time reflect the spatial distribution of carbon emissions; (3) Atmospheric emission of Xi'an will be 41.92 million tons at the end of 2022, a decrease of 2.80 million tons compared with that of 2020, but an increase of 0.33 million tons from 2021. The north of Xi'an and periphery of the central urban area are the main carbon sink loss areas, while the east of Xi'an and north foot of the Qinling Mountains are carbon sink growth areas.