利用夜间灯光数据进行长时期社会经济问题研究时，需要对数据饱和校正，从而得到可信可靠的研究结果。针对不变区域法在长时间序列夜间灯光数据饱和校正时假设不变区域数据不随时间变化以及未区分数据饱和部分和未饱和部分的不足，本文提出首先对数据年际校正，再以NDVI数据为辅助进行饱和校正的方法。年际校正时准确定义了基准区域和基准年份，饱和校正过程中分别对不同城市聚类分区构建校正模型。研究发现，夜间灯光数据包括未饱和部分和饱和部分，饱和阈值为30；两部分数据亮度值与相应无饱和数据亮度值的函数关系不同，未饱和部分符合线性模型，饱和部分符合指数模型；区分不同城市聚类分区进行饱和校正十分必要，尤其是大范围区域数据饱和校正；以NDVI足迹数据为辅助，运用指数模型对饱和部分数据校正后，数据值域增大，空间异质性增强，与区域GDP拟合程度改善，很好地消除了由于卫星传感器设置特性产生的饱和效应，得到更好反映人类社会经济活动强度和空间分布特征的长时间序列饱和校正夜间灯光数据。文中得到的年际校正和饱和校正模型可以不做参数调整而直接运用，校正方法适用性较强。

Night Time Light (NTL) data have been verified to be a favorable proxy for socioeconomic activities. However, saturation correction is necessary to make the results credible and reliable when detecting the multitemporal socioeconomic changes by using time-series analysis of the NTL data. This study is aimed at presenting a new method for correcting the saturation effects of the Defense Meteorological Satellite Program-Operational Linescan System (DMSP-OLS) stable light images based on normalized differential vegetation index (NDVI) data.
First, the different regions and years are selected as references for conducting intercalibration, which is different from the conventional invariant region method. A TNDVI indicator, which shows a significant positive correlation with the Digital Numbers (DNs) of the NTL data, is built based on the original NDVI data after the intercalibration. Second, a K-mean value is utilized to divide the cities into four types and then correct the saturation effects based on the various characteristics of the NTL and NDVI data of various regions. Third, the saturation threshold of the NTL dataset is accurately identified during the saturation correction. Furthermore, the saturated and unsaturated portions are analyzed to construct a saturation correction model. Finally, the relationship between the sum of the NTL brightness and Gross Domestic Product (GDP) before and after the saturation correction is compared to verify the effect of this new saturation correction method.
In this research, unsaturated and saturated portions can be found in the NTL dataset, with a saturation threshold of 30, that is, 0-30 are unsaturated, whereas 31-63 are saturated. The functions between the DN values of the two portions and the corresponding original data are different, that is, the unsaturated portions comply with the linear model, whereas the saturated portions comply with the growth model. The various cluster partitions must be distinguished, and the saturation effects, especially the saturation correction for NTL dataset in large areas with remarkable regional differences, must be corrected. On the basis of the TNDVI data, the DN values of the NTL images after the saturation correction by using the growth model for the saturated portions are remarkable, that is, the DNs of F182013 are from 2.717 to 245.673 after the saturation correction, the spatial heterogeneity is enhanced, the fitting relationship with the regional GDP is improved, and the saturation effects caused by the satellite sensor set attributes are appropriately removed. Therefore, the data can appropriately reflect the intensity and spatial distribution of human socioeconomic activities without the saturation effects. Hence, the new saturation correction method is confirmed to be effective.