Urban Built-up Area Extraction From Logarithm Transformed NPP-VIIRS Nighttime Light Composite Data

Whether at regional scale or global scale,figuring out the exact boundary of urban built up area is not only a key basis for urban management,but also the foundation for researches on politics,economy,civilization and environment.Considering the rapidly development of technology,urban sprawl speeds up,which in return makes the urban extraction a popular and focus research issue.Fortunately,the origin,the foundation,the planning and the development of urban structure are all in a close relationship with human activity.While the nighttime light remote sensing data has been proven to be the source that intimately correlates to human activity degree.This gives a clue of extracting urban built up areas based on the nighttime light remote sensing data.However,until now there barely are researches applying the Suomi National Polar-orbiting Partnership-Visible Infrared Imaging Radiometer Suite(NPP-VIIRS)remote sensing data for urban extraction.This article systematically explored how to extract urban built up areas based on the NPP-VIIRS monthly composite remote sensing data.The contents and the results of this research are methodically as follows:(1)This research figures out the key problem of extracting urban built up areas from NPP-VIIRS composite remote sensing data.The study systematically compared the characteristics and features between nighttime lights from Defense Meteorological Satellite Program-Operational Linescan System(DMSP-OLS)and NPP-VIIRS monthly composite remote sensing data.Through analyzing the good and the disadvantages,as well as figuring out what the similarities and differences between these two dataset,this article discovered the key problem is the magnitude and spatial heterogeneous of NPP-VIIRS nighttime light radiance value.(2)Solved the principle issue about urban built up area extraction based on NPP-VIIRS nighttime light dataset.Considering its skewed radiance distribution,this article proposed the logarithm transformation for pre-processing.It is beneficial in two aspects.Firstly,it normalizes the skewed distribution in a certain extent and stabilizes the increasing rate for dependent variables.Secondly,it downscales the radiance magnitude,which makes the distribution much easier to be observed and interpreted.(3)This article compared the results from four urban built up area extraction methods including sudden jump of urban perimeter,Laplacian corrected Sobel edge detection,Nearest neighborhood statistics analysis and watershed segmentation applied on original NPP-VIIRS and the logarithm transformed NPP-VIIRS composite data.It is revealed that the logarithm transformation is capable to improve the extraction results,which improved Kappa from 0.54 to 0.82.(4)This article at last analyzed the advantages and disadvantages of all the four methods and provided some suggestions.According to the research,watershed segmentation achieved the best results.The shape of urban built up area extracted from the nearest neighborhood statistics is most similar to that of the evaluation database.The urban perimeter-sudden jump is not suitable for the urban built up extraction based on NPP-VIIRS nighttime composite data.