| The primary condition for the establishment of distributed model is the calculation of terrain attributes and feature extraction based on DEM.The extraction of digital river networks is the core task of digital terrain analysis.lt is an important problem in the field of geographic information system.lt is a long history that human understand the basin.Therefore. a large number of river networks extraction algorithms.based on runoff simulation,have been proposed.Although different algorithms have achieved definite results under certain terrain conditions, most of these algorithms have a common defect.It is only for a single mountain or plain area for analysis and extraction.It has a certain one-sided and subjective about description of water systems on real terrain.It does not consider the river networks discrepancy of hills and plain area, so it leads to large differences in the river networks and the actual situation.Drainage area threshold determines basin networks morphology, level and density characteristics.Most of the existing studies focuse the influence of drainage area threshold on single basin.Relationships between drainage area threshold and topography, geomorphology, geology, soil, precipitation and vegetation are rarely analyzed.A single drainage area threshold seriously affects the river networks distribution and the extraction accuracy of river networks in complicated situations of underlying surface.Most of the river networks extraction algorithms have high complexity and are serial implementation of the iterative algorithms,so they are time-consuming.The limitation of computational efficiency has become one of the bottlenecks in the application of river networks.One of the problems to be solved is how to improve the execution efficiency of river networks.Simultaneously,how will the automated extraction of river networks and flood routing model,in order to optimize the accuracy of flood routing. And,it is crucial to solve the difficulty of manual control of stream sediments in the arrangement of exploration,how will the automated extraction of river networks and geochemical investigation combined with.Therefore,the purpose of this paper is to study application of river networks in meteorology and geochemistry.with the starting point of solving these problems existing in flood routing and geochemical exploration.Through the analysis and optimization of the existing algorithms of the river networks extraction based on slope ronoff simulation,this paper describes the proposed adaptive terrain of high precision river networks extraction model and in the exploration of the flood routing and chemical application.DEM data and a variety of macro and micro terrain factors are used as research objects.Based on the statistical analysis of the fuzzy landform classification method, the terrain is divided into two types of mountain and plain. At the same time, according to the geomorphic characteristics, the optimal river networks extraction algorithm is proposed.The multiple flow direction heuristic information algorithm is applied to the plain region,and the multiple flow direction flooding algorithm is used to mountain area.According to the accuracy requirements of the adaptive terrain river networks extraction model and combining with different climate and underlying surface conditions,these relationships between the drainage area threshold and basin precipitation, topography and surface cover are studied.The drainage area threshold is set independently by using multiple linear regression equation between each factor and threshold.At the same time, according to the real-time requirements of river networks and ensuring the accuracy of the algorithm, the original serial algorithm is developed to be parallel in order to improve the efficiency combining with the computer parallel technology. Eventually,The river networks are applied to flood routing simulation.Double-layer asynchronous iterative of water flow allocation strategy is optimized.It effectively simulates the flood evolution process under the natural terrain.Furthermore,the sampling and arrangement method of automated stream sediment is implemented,combination of river networks and regional geochemical exploration.It reduces the data error caused by human,and it improves the accuracy and efficiency of stream sediment sampling.The main research results of this paper are as follows:1) A high-precision algorithm of extraction river networks based on terrain adaptionIn this paper,based on the statistical analysis of the fuzzy landform classification method, the terrain is divided into mountain and plain.The result of automatic division of landform is consistent with the pattern of undulating topography, and the non supervised classification method based on the best terrain feature space is reliable.The algorithm of triangular facet multiple flow based on maximum downslope gradient is proposed in order to better simulate water slope on the loose flow(short TFM_md).Firstly, four kinds of artificial simulated surfaces are constructed;secondly, the SCA calculation is used to evaluate the accuracy of TFM_md.The results show that the TFM_md algorithm is superior to the other six general flow direction algorithm in terms of flow direction and flow accumulation,and it has less residual distributionthe of SCA.Accordingly,TFM_md is appropriate and credible, and it can fully simulate overland flow process.According to the result of landform division,the multiple flow direction flooding algorithm for river networks(short MFDFN) is proposed. It is suitable for use in the mountain areas,choosing Zhanghe River as the research object.The paper has compared the MFDFN algorithm and other seven kinds of river networks extraction algorithms from qualitative spatial form and quantitative multi range buffer river networks evaluation.Experimental results show that the MFDFN algorithm can extract continuous wide certain river networks.It effectively reduces the parallel channel and pseudo channel.It is more consistent with the actual topography.Meanwhile,the buffer distance standard deviation of MFDFN algorithm is 0.58 in the ten buffer.Thus,the MFDFN algorithm has a good performance.Meanwhile,the multiple flow direction heuristic information algorithm for river networks(short MFDHN) is applied to the plain region,choosing Hanshui River as the research object.The paper adds the real networks to the MFDHN algorithm.The MFDHN uses heuristic preprocessing and TFM_md strategy. The river networks of Hanshui area are extracted by D8 algorithm and MFDHN algorithm. Experimental results show,as far as the spatial distribution,the MFDHN algorithm can extract continuous river networks.It effectively reduces the parallel channel and pseudo channel.Therefore,the high-precision algorithm of extraction river networks based on terrain adaption improves the accuracy,and it has clear physical meaning.2) Seting drainage area threshold based on the characteristics of DEMThe mean annual precipitation and vegetation cover data are used as samples. which are provided by the Regional Climate Center of Wuhan from 1975 to 2010 in Hubei province.Elevation data with scale of 1:50000 is used as the base data.Twenty sample regions of different underlying surface conditions are selected as the research objects.The influence of some factors,such as regional ground slope,precipitation and vegetation cover,are researched on drainage area threshold.In this paper,the comprehensive evaluation index of drainage area threshold and average slope,mean annual precipitation,vegetation cover is studied by multivariate regression analysis.A goodness of fit of the equation of linear regression with multiple condition regression analysis is 0.837.The significant test of equation is 0.9998.So,the whole regression equation is significant.According to the model of catchment area threshold, some typical basins are selected to test the regression equation.Experimental results show that the river networks of simulation are consistent with practical under different terrain, vegetation coverage and precipitation.Therefore,seting drainage area threshold based on the characteristics of DEM data is a reasonable and feasible method using multiple linear regression model.3) Parallel algorithm of raster river networks extractionHaving analyzed the feasibility of river network extraction on GPU implementation,the parallel algorithm of river networks extraction based on CUDA is put forward.The serial algorithm is very time-consuming on flow distribution coefficient and flow accumulation. Data partitioning and asynchronous transfer strategies are designed to implement parallel algorithms. In order to evaluate the performance of the algorithm,the computations of different size DEM data take time between sequential implemented on CPU and parallel implemented on GPU in Zhanghe River. Experimental results show that the relationships between the flow calculation,river networks extraction and grid size are quadratic polynomial. When the amount of data is reached 5160X 7260,the speed-up ratios of flow calculation is 28.4 times,and the speed-up ratios of river networks extraction is S6.8. So,as the amount of data increases, the speed-up ratios increases.The parallel algorithm can improve the speed of river networks extraction.It can be used for real-time interactive feature extraction. 4) Research of flood routing simulation based on river networks Combined with actual terrain information,one arithmetic of river networks has to been used for simulation of flood evolution process. TFM_md algorithm is used to optimize the water allocation strategy for double-layer asynchronous iterative. It effectively solves the problem of simulation accuracy of flood evolution caused by single-flow direction homogenization iteration. The adaptive selection mechanism of iteration step is established. Cao River is selected as the experimental area.Simulation results of different rainfall times are compared and analyzed. Experimental results show that the algorithm is sensitive to terrain and water flows to low-lying areas. Meanwhile,the overall differences of the algorithm and FloodArea algorithm are not generally remarkable.compared with the iterative homogenization algorithm. So.the algorithm can effectively simulate the flood evolution process.lt can obtain real-time flood submerging range and water-deep distribution.lt also can provide the basis for flood control command, risk map making and risk assessment.5) Methods of sampling and layout on automated stream sedimentIn this paper, computer technology has been combined with geochemical explorarion.Through theoretical analysis and experiment,the method of sampling and layout on automated stream sediment is put forward,starting from the point of view of the geochemical work arrangement and combining high-precision algorithm of extraction river networks based on terrain adaption.The purpose of selecting the experimental data is to compare and analyze the difference between the sampling points of drawing in hand and the automatic layout of stream sediment. The contour curvature method is used to obtain the valley locations. Experimental results show that automated extraction of river network is consistent with the hand drawn.The difference of number and location of sampling points of stream sediment by the two methods is very small.It can meet the actual production requirements.Therefore,it improves the efficiency of stream sediment sampling. The Innovation of this paper is mainly embodies in: 1) Using TFM algorithm to construct the inner surface of the pixel and optimization of flow distribution weights based on MFD_md algorithm,the algorithm of triangular facet multiple flow based on maximum downslope gradient is developed (short TFM_md).The algorithm is divided into two levels.It include two parts:flow routing within a cell and flow distribution to neighboring cells.TFM_md algorithm can ensure the accuracy of the original terrain.The influence factors of water flow allocation strategy are considered and it has a clear physical meaning.2) Adaptive high precision river networks extraction algorithm based on terrain is constructed. According to the automatic classification of different landform(mountain and plain),the algorithm can adaptively select the river networks extraction model.It effectively reduces the parallel channel and pseudo channel, and improves the extraction accuracy of river networks.3) In this paper, Automatic selection of model of drainage area threshold based on multi factor regression equation is constructed.It can quickly determine the optimal threshold value of the catchment area.It provides reference for runoff yield and concentration of distributed hydrological model.4) Using TFM_md algorithm to optimize the water allocation strategy for double-layer asynchronous iterative,flood routing model of multiple-flow direction double-layer asynchronous iterative in the river network is developed.The model can effectively simulate the natural terrain conditions of river overtopping flood evolution process. |
PDF Full Text Request