Multi-source Remote Sensing Data Fusion And Synergistic Technology For Dynamic Changes Of The Estuarine And Coastal Water Environment

Estuary is a transition zone of land-sea interaction,where the ecological environment is fragile,the dynamic environment is complex,and the water environment is characterized by strong changes.The development of satellite remote sensing observation technology has provided a strong and favorable means to monitor the changes of water environment in estuarine coast.However,there is not yet a single data source on board that is specifically designed for monitoring water environment in estuarine coast,i.e.,a single data source that satisfies spatial resolution(<30 m),spectral resolution(<20 nm),and temporal resolution(<1 h)at the same time.To show the advantages of multi-source satellite data integration,this paper takes the near-shore suspended sediment of estuaries,dynamic environment and ecological disasters as the observation and application objects,and develops the methods of combining remote sensing,ground truth and numerical simulation of multi-source heterogeneous data.More specifically,through machine learning,feature recognition and tracking,spatio-temporal fusion and many other technical studies,remote sensing detection capability was improved and remote sensing information potential was deeply explored.This study lays the foundation for the realization of an estuarine coastal system observation system based on multi-source heterogeneous data,main research contents and results are as follows:(1)Cross comparison and validation of the consistency of multi-source remote sensing data applied to estuarine coastal water environmentThe applicability of multi-source satellites on estuarine and coastal water observation was studied and analyzed,for the purpose of providing the data support and theoretical basis for the integrated observation of remote sensing data.Among the three high spatial resolution land observation sensors used in this study,Landsat-8/OLI(L8/OLI),GF-1/WFV,and Sentinel-2/MSI(S2/MSI)all had good agreement with the ocean color sensor GOCI with the R~2>0.9.The analysis of signal-to-noise ratio indicated that the L8/OLI and GF-1/WFV were able to identify Suspended Particulate Matter(SPM)concentration variations when over 0.2 mg/L,while S2/MSI was only able to identify SPM concentration variations when over 0.5 mg/L.Based on the coefficient of variation,The spatial resolution impact analysis showed that the information within the 10 m resolution image is about 1.2�1.4 times than that within the 500 m resolution image in the same area.(2)Improved remote sensing inversion algorithm for SPM concentration in high turbidity estuarine watersOn the basis of the Semi-Empirical Radiative Transfer(SERT)model,the improved SPM inversion model(SERT?r),which is based on the band ratio,has been developed and constructed.The inversion accuracy of SERT?r is better than that of common SPM inversion models and is applicable to waters with different turbidity.There is good agreement between the derived SPM and the measured SPM,with R~2 of0.85 and MAE of 112 mg/L.(3)Improved multi-source and spatial-temporal fusion model using deep learning techniquesThe 200 pairs of high-and low-resolution optical image data of the estuary were trained to construct a spatial scale conversion model based on Super-Resolution Convolutional Neural Network(SRCNN).Combined with SRCNN and the Spatial and Temporal and Temporal Adaptive Reflectance Fusion Model(STARFM),remote sensing products at high temporal and spatial resolutions were reconstructed.The validation on the simulated data indicated that the prediction error of SRCNN-STARFM in SPM retrievals in the horizontal motion was reduced by 1.2%,and that in the vertical motion was reduced by 0.6%.(4)Proposed methods for analyzing hydrodynamic variation patterns based on multi-source heterogeneous observation dataThe hourly ocean color remote sensing data from GOCI were used to analyze the variations of SPM distribution before and after four typhoons,particularly for Typhoon Soulik.By combining the remote sensing images with in-situ measurements from ground stations,meteorological re-analysis data,and numerical simulation data,the results of the comprehensive observations showed that the SPM concentration at the Yangtze Estuary increased rapidly during the typhoon transit,the high SPM concentration continued for 1�2 days after the typhoon dissipated.Under the influence of the typhoon,the average significant wave height outside the Yangtze Estuary increased by 2.1 times.For the lower mouth of the Yangtze Estuary,about 53%of the SPM concentration and 10.0%of the bottom shear stress were directly originated from wind driven wave during the Soulik typhoon.Meanwhie,the Maximum Cross Correlation(MCC)technique was developed to solve the failure of flow field retrievals in the mouth of river.Validated by the measured data,the results show that the inverse flow field optimized by the multi-band TOA reflectance,dimensional switching and reciprocal filtering techniques had the highest accuracy.The error of flow direction and velocity were 24�and 0.14 m/s,respectively,which were better than the numerical simulation results.(5)Proposed a synergistic analysis method of optical-microwave multi-source remote sensing data for the evolution of small-scale algal bloomTo compensate for the reduced temporal resolution of optical images due to cloudy weather in the coastal area,a synergistic analysis study of optical-optical-microwave remote sensing data was conducted.According to the signal characteristics in both optical and microwave images,a high-precision classification and extraction method of algal marsh was proposed with the classification accuracy over 99%.Meanwhile,the overestimation of algal bloom area due to the mixing effect of image elements was analyzed.Based on the high temporal resolution remote sensing data,high frequency observation of the growth and elimination processes of Enteromorpha prolifera bloom in the Yellow Sea were obtained.In summary,this study can provide new technologies and methods for the construction of a multidimensional monitoring system for estuarine coastal water environment,as well as the support of high-precision data products,which highly improves the capability of monitoring estuarine water environment elements,estuarine surface flow field,and near-shore ecological disasters.