LAUSR

Background noise of nighttime light (NTL) imageries derived from the Visible Infrared Imaging Radiometer Suite (VIIRS) day/night band (DNB) adversely affected the accuracy in investigating socio-economic activities, which needed to be reduced. The threshold method was often used in noise reduction, and the critical issue of which was how to determine the threshold. The monthly composite VIIRS NTL data and electric power consumption (EPC) of 14 provinces in southern China between 2013 and 2018 were used in this letter, which aimed to figure out the optimal threshold for denoising VIIRS NTL data and estimate monthly EPC using denoised NTL data. The results show that: 1) monthly composite VIIRS DNB NTL data is reliable in estimating monthly EPC with high accuracy; 2) it is reasonable to determine the optimal threshold according to the $R^{2}$ of fitting; 3) the optimal denoising threshold is not exactly the same in each month; and 4) 0.8 is recommended to be a uniform threshold for all months if needed.