| Marine human activities(overfishing,destructive fisheries,etc.)are one of the factors leading to the degradation of coral reefs,and the current long-term data on marine human activities are relatively lacking.The Operational Linescan System(OLS)of the Defense Meteorological Satellite Program(DMSP)provides a rich historical archive of night-time light remote sensing images covering a wide range of ocean areas,with potential for indicating human activities that dominated by fishing activities in coral reefs region.However,due to the systematic radiation error of DMSP/OLS night-time light remote sensing time series images,it cannot be directly used for marine human activities research,and its indication effect on marine human activities has not been verified.Research of impact of human activities indicated by images on coral reefs is also relatively scarce.In response to the above problems,this study firstly proposes an intercalibration model based on random forest noise reduction,and uses DMSP/OLS average visible light images to construct a night-time light remote sensing time series images with consistent radiation in a large sea area.Then,the fishery data is used to validate the indication effect of night-time light remote sensing images on fishery activities.Based on this,the influence of marine human activities in coral reef areas indicated by night-time light remote sensing images on coral reef ecology is explored.The temporal and spatial distribution characteristics of human activities in the coral reef area of South China Sea were extracted to provide data support for the management and research of coral reefs in the South China Sea.The main research contents and conclusions are as follows:(1)A method of constructing ocean night-time light remote sensing time series images is proposed,which solves the problem of DMSP/OLS average visible light image radiation error in sea area.First,used the average visible light images,median filtered images,variance filtered images,and bathymetric data images to synthesize feature image,and trained a random forest classifier through the feature image to remove the noise to obtain effective pixels in the classification of feature image.Then cross-validation was used to select a fitting function,and a stepwise relative correction model was established to reduce the systematic error of the denoised image.The results showed that the average overall classification accuracy of the random forest noise reduction model was96%.The stepwise intercalibration model based on random forest noise reduction can effectively reduce the systematic radiation error of the image.Compared to the original average visible light image,the radiance error was reduced by about68.2%.Accordingly,this study established a long-term(1992-2013)night-time light remote sensing image dataset of large-scale sea areas with consistent radiation.(2)The indication effect of ocean night-time light remote sensing time series images on marine human activities is discussed.Investigate the correlation between the fishery data in the exclusive economic zones of Japan,Malaysia and Thailand and the total light index of night-time light remote sensing image.The results showed that there was a significant correlation between the fishing catches in the exclusive economic zones of Japan and Malaysia and the total light index of night-time light remote sensing images(R=0.925,p<0.01).There was a correlation between the number of small fishing boats in Thailand’s exclusive economic zone and the total light index(R=0.5,p<0.05),in other words,night-time light remote sensing images can effectively indicate marine human activities dominated by fishery activities.(3)The influence of human activities on coral reef ecology indicated by ocean night-time light remote sensing time series images is discussed,and the temporal and spatial distribution characteristics of human activities in coral reef areas in the South China Sea are extracted.The results of correlation analysis,buffer analysis and standard deviation ellipse analysis showed that:a)human activities indicated by the night-time light remote sensing images had a significant negative impact on coral reefs,and the coverage of living corals was significantly negatively correlated with the human activity intensity index(R~2=0.627,p<0.01);b)the dramatic degradation of coral reefs in the Xisha Archipelago may be linked to the sharp increase in marine fishing activities from 2006 to 2008;c)from 1992to 2013,the intensity of human activities in the coral reef area of the South China Sea showed an upward trend,and the scope of its influence also showed a trend of expansion.Specifically,human activities in the waters near Dongsha Atoll and Zhongsha Atoll increased rapidly around 2006,and human activities in the coral reef area of the northern Nansha Archipelago increased significantly around 2011. |
