Research On Parallel Computing And Remote Sensing Data Generation Method For Distributed Hydrological Simulation

As an important way to understand the water cycle,hydrological simulation,after more than half a century of development,has entered the stage of conceptual lumped model from the initial stage of empirical black-box model,and is now in the stage of distributed hydrological simulation.Distributed hydrological simulation has a large amount of calculation and requires many data parameters.More and more researchers have proposed a parallel computing method for distributed hydrological simulation,which combines remote sensing data to obtain hydrological parameters and vegetation parameters of hydrological simulation.Hydrological simulation.In this paper,a parallel computing system is designed based on the Dis VOM hydrological model,and the spatial-temporal fusion and spectral fusion techniques are used to reconstruct remote sensing images to solve the problem of lack of data.The main research contents are as follows.(1)In this paper,by analyzing and comparing the existing parallel computing methods for hydrological simulation,the method based on task parallelism is used to rasterize the Poyang Lake watershed in Jiangxi Province,and the DAG graph of computing tasks is constructed according to the flow direction relationship and hierarchical relationship between the grids,and the resource management is coupled at the same time.The system PBS and Dis VOM hydrological models improve the efficiency and stability of system computing resource scheduling.This method has carried out a distributed hydrological simulation parallel computing experiment in the Poyang Lake Basin.The results show that the distributed hydrological simulation parallel computing system constructed based on this method can achieve parallel computing with a scale of over one million computing tasks,and in terms of computing speed,Using high-performance computing clusters,parallel computing is greatly improved.Due to the lack of relevant vegetation data,meteorological data and other parameters.The calculation parameters used in the experiment are all artificially set,and the experimental results cannot accurately reflect the hydrological simulation of the real environment.(2)In order to solve the problem of lack of parameters,this paper uses the existing spatiotemporal fusion algorithm to reconstruct remote sensing images to obtain images with high spatiotemporal resolution,so as to provide data support for the subsequent parameter inversion.In this paper,several algorithms such as EDSR,FSDAF,Fit-FC,SFDAF,SPSTFM and VIPSTFM are selected to reconstruct the spatial-temporal fusion image of the Landsat image and MODIS image in the Poyang Lake area for analysis.Analyzing the experimental results,for the reconstruction of remote sensing images of the Poyang Lake Basin,the existing spatiotemporal fusion algorithm has a better effect on the reconstruction of the ground objects with insignificant changes and short time intervals.not effectively.(3)In order to solve the problem of unsatisfactory effect of spatial-temporal fusion reconstruction of remote sensing data in areas with drastic changes in spatial information.In this paper,the synthetic aperture radar image(Synthetic Aperture Radar SAR)will be used to provide high-precision spatial information,and the SAR image will be spectrally reconstructed using the MODIS image based on the GAN network idea.Based on GAN network,this paper proposes a new network SARGAN for spectral fusion to obtain high-precision images with spectral information.According to the experimental results,it can be seen that the fusion method proposed in this paper is effective,and the spectral information of the reconstructed image is better than the results of all spatiotemporal fusion algorithms used in this paper.(4)Since the imaging principle of SAR data is obviously different from that of optical images,and it does not contain spectral information,the reconstruction effect of the spatio-temporal fusion algorithm is not good due to the large water area and drastic changes of Poyang Lake.In this paper,the reconstruction results of the two are spliced together.For the reconstruction results of the water part,use the spectral fusion results,and use the spatiotemporal fusion results for areas such as woodland and buildings.For the extraction of waters,this paper analyzes and compares the methods of water extraction from SAR data,and uses multi-threshold combined with decisionmaking method to extract waters from SAR data in the Poyang Lake Basin.The extracted water area is used as a mask to replace the water part of the high-resolution multispectral data generated from the optical data spatiotemporal fusion in the previous study and the SAR domain MODIS data to improve the accuracy of spatiotemporal fusion.The experimental results show that the high-precision spatial information provided by SAR data can play a great role in the improvement of spatial-temporal fusion water reconstruction.