Spatio-temporal pattern evolution of energy consumption carbon emissions at the city and county levels based on the population-kNDVI correction of nighttime light from 2000 to 2020

• kNDVI and population density correction for nighttime lights. • Integration of DMSP-OLS and NPP-VIIRS to construct long-term series of nighttime light. • Using the GTWR model to establish a carbon emission estimation model based on night lights. • Multi-scale analysis of spatio-temporal changes in carbon emissions. Energy consumption increases carbon emissions,exacerbating climate change. The study of the spatio-temporal evolution of carbon emissions contributes to the formulation of carbon reduction policies.The characteristics of the spatio-temporal evolution of carbon emissions are important for the precise formulation of carbon reduction strategies. However,carbon emissions estimates focus primarily on provincial level. Due to insufficient data and deviations in energy statistics, accurate carbon emissions estimates at these scales remain challenging at city or county level.Human activities can be recorded through nighttime light, which is useful for estimating carbon emissions. Therefore, how to process nighttime light to enhance the accuracy of carbon emissions estimates is a critical issue.This study combines nighttime light with population density and kernel NDVI to propose PKANTL for calibrating nighttime light,Bayesian optimization of long short-term memory networks is used to construct long-term sequence nighttime light.Using a geographically weighted regression model to link carbon emissions with PKANTL to estimate carbon emissions, generating the spatial distribution of multi-scale energy consumption carbon emissions from 2000 to 2020.The research results are as follows: the R²of BO-LSTM fitting DMSP-OLS and NPP-VIIRS is 0.9422, demonstrating high accuracy. PKANTL can effectively process nighttime light, highlighting human activities.The average R 2 of carbon emissions simulated by GTWR is 0.865, which is more accurate than linear regression and quadratic polynomial.Carbon emissions from energy consumption exhibited an growth pattern during the period from 2000 to 2020.Carbon emissions are clustered,High carbon emission areas are predominantly located in the Beijing-Tianjin-Hebei region in the central part of China and the eastern coastal areas,with a predominance of slow growth at city level.At county level,carbon emissions are also clustered,primarily in the e Yangtze River Delta, Pearl River Delta, and Counties of some provincial capitals, exhibiting mostly moderate growth.The research results demonstrate the spatio-temporal distribution of carbon emissions from energy consumption in China, providing data support for achieving carbon reduction targets.