Abstract: The Loess Plateau, a typical ecologically fragile region and energy-intensive development zone in China, exhibits spatial heterogeneity in carbon emissions that has yet to be precisely characterized. To overcome the limitations of coarse spatial resolution in traditional statistical data, we integrated DMSP/VIIRS nighttime lights, GLC_FCS30 land use grids, meteorological, and socioeconomic data to construct a distributed carbon emission estimation model combining night light correction with energy coefficient weighting, generating a county-level carbon emission dataset with 1 km spatial resolution. We used standard deviation ellipses, spatial autocorrelation, and geographic detector methods to reveal the spatiotemporal variations and driving factors of regional carbon emissions from 2010 to 2020. Results showed that total carbon emissions in the study area decreased from 1.89×10⁸ t to 1.74×10⁸ t between 2010 and 2020. The center of gravity for carbon emissions shifted northeastward by 196 km, with enhanced spatial clustering. Carbon emission intensity was highest in construction land (63.42 t·hm^-2), significantly exceeding that of forest and farmland. The afforestation and grassland restoration reduced regional emissions by 1.30×10⁷ t, while urban expansion contributed about 1.78×10⁷ t of new emissions. Geospatial analysis indicated that urbanization rate (q=0.3812) and economic development level (q=0.2976) were dominant factors shaping spatial carbon emission differentiation, while the interaction between energy structure and land use exhibited a significant nonli-near enhancement effect (q=0.4011). The distributed accounting method, incorporating nighttime light correction and energy coefficient weighting could reliably capture the spatiotemporal patterns of county-level carbon emissions. This approach would provide scientific evidence and data support for coordinating energy development with ecological conservation, optimizing land use structures, and formulating differentiated emission reduction policies in the Loess Plateau region.

Key words: multi-source remote sensing data, land use carbon emission, spatiotemporal evolution, geographic detector, Loess Plateau