| Earth observation satellites provide a lot of important information for industrial production and daily life, with the capacity of long-term and large-scale earth system monitoring. Remote sensing data has been widely applied to cloud research, mainly for two reasons: on the one hand, cloud is an important interference factor in many remote sensing applications, such as the monitoring of the ground, so removing the clouds is a prerequisite for further applications; on the other hand, clouds play an important regulatory role in the earth-atmosphere system, in order to accurate prediction of the climate, the construction of a numerical model requires a large number of accurate cloud parameters with regional or global coverage. The contents of this paper are: first, we conducted an urban-scale air quality assessment application with four empirical models regressed by PM2.5 and AOT derived from MODIS L1 B within which clouds have been removed by the cloud mask from cloud detection products; secondly, we retrieved high spatial resolution cloud fraction dataset using MODIS cloud detection products, and through a series of comparative analysis with surface observations of cloud fraction, we obtained high quality cloud fraction dataset not only for studying the climate change, but also for evaluating the MODIS cloud detection products which provides further research direction in proposing a cloud detection algorithms; Finally, we analyzed the MODIS cloud detection tests. The main conclusions of this paper are listed as follows:1. Four empirical models which are used to investigate the relationship between AOT data and PM2.5 mass concentration are obtained by regression analysis and the accuracy of them are R2=0.818, R2=0.750, R2=0.699 and R2=0.629 respectively. The quadratic model has significant potential to enhance air quality monitoring at urban scale.2. We use these models to retrieved PM2.5 concentration from MODIS AOT on 11 th of October 2012 and then compare with the PM2.5 concentration from ground-based in that day, 50%, 46.4%, 46.4% and 39.3% of stations are within the expected errors respectively by four models.3. Based on the time series analysis of morning(Terra) and afternoon(Aqua) cloud fraction derived from MODIS cloud detection product, we found that cloud fraction trends reduced slightly and more clouds during afternoon than the morning. A comparison between satellite and surface observations of daily cloud fraction presents a good relation with correlation coefficient of 0.878 for the period of 2012.4. Analyses of monthly mean cloud fraction between satellite and surface observations show larger discrepancy in winter, which can be explained by cloud detection procedures when dealing with bright surface, like desert or snow cover.5. The statistics test is carried for six selected study areas and the result indicates that the correlation between satellite and surface observations of cloud fraction increases clearly in the north region after removing winter data, especially in the northeast forest area, while the correlation coefficients in southern China do not show significant changes due to nearly no snow in winter.6. Based on cloud detection test, we found the use of bright temperature differences test can effectively suppress the mistaken which treated the bright desert or sparse vegetation cover area as clouds. |
PDF Full Text Request