ABSTRACT

Monthly Visible Infrared Imaging Radiometer Suite (VIIRS) Day-Night Band (DNB) composite data are widely used in research, such as estimations of socioeconomic parameters. However, some surface conditions affect the VIIRS DNB radiance, which may create some estimation bias in certain regions. In this paper, we propose a novel normalization algorithm for VIIRS DNB monthly composite data. The aim is to normalize VIIRS radiance, collected under different surface conditions, to a reference point, so that the bias is reduced. The algorithm is based on the utilization of stable lit pixels as a reference and a nonlinear regression algorithm, to match un-normalized data to the reference data. Experimental results show that the algorithm could improve correlation (R²) between the total sum of nightlights (TOL), electric power consumption (EPC), and gross domestic product (GDP) at both a global and local scale. The algorithm could significantly diminish the seasonal component of un-normalized nightlights radiance caused by snow. The intensified nightlights radiance in sandy regions could also be reduced to a more reasonable range in comparison with other regions. Visual inspection shows that the brightness of snow-affected and sandy regions was strongly reduced after undergoing normalization.

摘要

每月可见红外成像辐射计套件（VIIRS）昼夜波段（DNB）复合数据广泛用于研究，例如社会经济参数的估算。但是，某些表面条件会影响VIIRS DNB的辐射度，这可能会在某些区域产生一些估计偏差。在本文中，我们为VIIRS DNB月度复合数据提出了一种新的归一化算法。目的是将在不同表面条件下收集的VIIRS辐射归一化到参考点，以减少偏差。该算法基于利用稳定点亮的像素作为参考和非线性回归算法，以将未归一化的数据与参考数据进行匹配。实验结果表明，该算法可以改善相关性（R ²）在全球和本地范围内的夜灯总数（TOL），电力消耗（EPC）和国内生产总值（GDP）之间。该算法可以显着减少雪引起的非标准化夜灯辐射的季节分量。与其他地区相比，在沙质地区增强的夜光辐射也可以降低到更合理的范围。目视检查表明，在进行归一化处理后，受雪和沙质影响的区域的亮度大大降低了。