Enhancing near-real-time gridded carbon emissions estimation with fine-granular electricity data and meteorologically normalized NO2 fields

Timely and spatially explicit near-real-time (NRT) carbon emissions data are essential for effective climate action in rapidly changing city environments. However, existing NRT datasets, such as Carbon Monitor, generally rely on generalized proxy variables with coarse spatial resolution, which introduce uncertainties at the (sub)city scale and limit their usefulness for city policy. To address this gap, we present the Multi-data-driven Carbon Accounting Network (MDCAT), a method for estimating daily NRT carbon emissions at 1 km resolution across the industrial, power, residential, and transportation sectors. Using Chengdu (a megacity in China) as an example, MDCAT integrates fine-granular electricity consumption data to estimate emissions for 35 industrial sub-industries and the residential sector, and employs meteorologically normalized NO2 (MN-NO2) fields as a proxy for traffic emissions. Cross-validation results showed that electricity usage was a reliable proxy, with Mean Absolute Percentage Error (MAPE) of 6.8 ± 3.1% for industrial and 2.3 ± 1.5% for residential emissions. Furthermore, MN-NO2 correlated strongly with city traffic NOx emissions (R = 0.83), a key co-emitter of CO2. Compared with existing products, MDCAT better captured abrupt, event-driven emission changes. For example, it identified a 54.6% drop in industrial emissions during the 2022 power curtailments and a 16.8% traffic reduction during the Chengdu 2021 International University Sports Federation (FISU) World University Games, whereas Carbon Monitor-Cities (CM-Cities) reported only marginal responses of 2.1% and 2.4%. These results demonstrate that combining high-resolution electricity data with MN-NO2 fields can substantially improve NRT city carbon accounting and support timely city emissions tracking and policy assessment.