夜间灯光遥感数据直接反映了人类活动的时空分布，因此被广泛应用于城市研究、社会经济活动监测等领域。DMSP/OLS数据提供了1992-2013年的历史夜间灯光观测，且卫星过境时人类活动相对活跃，因此在利用灯光数据的研究中该数据有着不可替代的地位。然而，由于受多种因素的影响，DMSP/OLS夜间灯光数据存在着三个主要的问题：年际不连续、过饱和效应、以及溢出效应，导致降低城市范围确定、相关社会经济指标估算的准确性。尽管目前有很多研究致力于解决这三个问题，但已有的研究仍然存在着一定的问题，尤其是后两个问题还未得到较好的解决，因此目前还缺乏可广泛使用的、高质量长时序的夜间灯光遥感数据。

本研究针对DMSP/OLS数据存在的上述问题，分别提出过饱和效应纠正方法和溢出效应纠正方法，以获得高质量、长时序的夜间灯光遥感数据。对于过饱和效应，我们提出了一种借助已有的无过饱和问题的DMSP/OLS辐射校正数据，结合对数模型以及双年调整的方法（logarithmic model with double-year adjustment , LMDA）对该问题进行校正。对于溢出效应，我们首先定量评估了灯光数据的溢出距离，在此基础上提出了一种无需借助任何辅助数据的自适应的溢出校正方法（self adjusting model, SEAM），并将其用于DMSP/OLS稳定灯光数据。本研究在中国区域对所提出方法进行检验，同时与已有的方法进行对比。最后，本研究将上述两种新方法结合已有的年际校正方法，在中国区域构建了一套完整的高质量DMSP/OLS时间序列数据，并从社会经济指标和能源消耗方面对该数据进行评估并得到以下结论：

（1）LMDA方法具有较好的重建饱和区域光强分布的能力。通过充分利用所有可用的真实灯光信息，包括DMSP/OLS辐射校正数据和稳定灯光数据，LMDA方法在去除过饱和问题方面更加可靠和准确。

（2）在中国区域内，NPP/VIIRS数据的溢出距离大约为1.7km，DMSP/OLS数据的溢出效应大约为3.5km。

（3）SEAM模型可实现对于溢出效应的去除，该模型校正后的结果与NPP/VIIRS影像更为相似，像元异质性被提高，同时对于城市提取的结果也更加准确。

本研究提出的校正模型能够有效地解决过饱和效应和溢出效应，且与已有的方法相比更加简单有效。结果表明，与原始数据相比，本研究构建的高质量的DMSP/OLS时间序列数据与国内生产总值和电力消耗具有更好的相关性，能够更准确地地反映社会经济和能源消耗的变化，有更高的应用价值。

In recent years, the night time light data has been widely used in urban studies, especially in the research of urban extraction and estimation of social-economic indexes for its ability of reflecting human activities. Thanks to the long data accumulation and appropriate passing time corresponding to the major human nocturnal activities, DMSP / OLS data plays an irreplaceable role in the study of historical human activities reflected by artificial lights. However, the DMSP/OLS lights product suffers from three main problems due to imperfect sensor design,  effect of atmosphere and other factors, i.e.inter-annual inconsistency, saturation and blooming effect which will affect the accuracy of urban extraction and the estimation of the social-economic indexes. The three problems have been recognized for a long time and great efforts have been made to mitigate their adverse effects addressed by some researches. However, so far, a robust and widely accepted method to alleviate the adverse effects does not exist, especially for the last two problems, which challenges the improvement of data quality of stable optical products. Therefore, there is still a lack of widely used night time light data with high quality and long time series.

This study aims to develop robust and reliable methods to exclude the saturation and blooming effect of the DMSP/OLS night time light data. For the saturation effect, we proposed a logarithmic model with double-year adjustment (LMDA) method by using DMSP/OLS’ discrete radiance-calibrated data to correct saturation effect in the annual stable light data.. For the blooming effect, we first measure the blooming distance of NTL data and explore the contributions of three main factors on it. Then, we developed a self-adjusting model (SEAM) based on spatial response function (SRF) to correct the blooming effect without using other ancillary data and tested the SEAM model to correct blooming effect in China. With the proposed model, we reconstructed the high-quality DMSP/OLS time series data in China with the existing inter-annual correction method, and evaluated the data using socio-economic indicators and energy consumption. Finally, we came to the following conclusions:

(1) LMDA can effectively reduce the saturation in the original DMSP/OLS images. By using all the available real light information, including the DMSP/OLS radiance-calibrated data and stable light data, LMDA method is more reliable and accurate in terms of excluding saturation.

(2) In China, the blooming distance of NPP/VIIRS data is about 1.7 km and for the DMSP/OLS data, it is about 3.5 km.

(3) The SEAM model can largely remove blooming effect in the original DMSP/OLS data and enhance its quality. The results is more similar with NPP/VIIRS image, and the variation of DN value within the urban area has been improved. Meanwhile, the city extraction results are more accurate. 

(4)  The methods proposed by this study can effectively correct the saturation and blooming effect of the DMSP/OLS nighttime light data. Compared with the existing method, our models are more simple and effective. The results show that the high-quality DMSP/OLS time series data has better correlation with GDP and power consumption compared with the original data. The reconstructed data thus can reflect the changes of social economy and energy consumption more accurately and has higher application value.