| Precision information of soil attribute is necessary for environmental management, monitoring, modeling, precision farming and other work. Currently available soil information is the map of soils obtained from traditional soil survey, the information on the type of soils in the polygons of the map is limited and discrete, and may even integrate other types. It is difficult to meet the social demand for high precision information on soil. However, in the previous studies of soil mapping, a huge collections of soil sampling data including soil nutrients, soil type and soil forming environment have been done, and many conventional soil maps have attracted the involvement of soil experts. These valuable historical information may be helpful to establish soil landscape model and be fully used in the acquisition of the precision soil information. So, we could not only reduce research costs but also improve efficiency and accuracy of research. However, accessing to information from soil is often realized by geospatial analysis.To improve the accuracy of soil information in the results of traditional soil mapping, four areas were choosed to explore, that is extracting and integrating of soil environmental factors, design and optimization the layout of soil sampling, soil-landscape modeling and improving, summarizing and displaying comprehensive soil information, using geospatial analysis based on the theory of geostatistics and soil-landscape modeling in this study. Zhongxiang City in Hubei Province, for example, was chosen to studied the design of soil sampling based on soil-landscape relationship, soil sampling optimization based on the layout of the road network, the prediction of soil organic matter based on the terrain units, the conclusion is as follows:(1) Relationship between soil and landscape of the study area in the case without human interference is fixed, so by comparing the similarity of soil landscape correlation matrix determined by the corresponding sampling points in different sampling programs, whether there is a fixed relational schema of the relationship between soil and landscape in the study area.can be checked. 8 topographical factors and 5 conventional nutrients in the study area A were used as investigating data for the study. With the application of spatial scalability and spatial statistics technology, the topographic factors were devided into five levels. Pearson correlation analysis was applied to the samples of each partition, and the similarity of Pearson correlation coefficients between topographic factors and soil nutrients in different sampling condition were analysed. The similarity between the Pearson correlation coefficients for the terrain factors and the soil nutrients under different sampling systems reaches over 99%, indicating that there is a fixed relationship pattern in the study area. The A sampling program which integrated a variety of environmental factors and determined based on topographical factors classification has a higher precision prediction, and its relationship model could be taken as a fixed relationship to assess the reasonableness of the remaining sampling scheme. So, the method is feasible. In this section, three types of traditional random sampling method based on grid have been designed. The relationship pattern obtained from sampling program C is the closest similarity to the one from sampling program A. It implied that a relational schema could be obtained from the sampling program C which contains 3661 sampling points, and it is more reasonable than other sample design.(2) The development of urbanization makes the regional transport network more developed and stable, which offering convenient traffic conditions for soil sampling in the field. The eastern region of Zhongxiang City in Hubei Province(research area B) was chosen for the study in this part. 5 types of sampling scale were set in soil sampling based on the road network. During the sampling process, the sampling data may contain errors due to a number of factors, and the spatial layout of sampling points may not be very reasonable possible. They will affect the accuracy of the final results of the study. Therefore, the sampling data with errors need to be eliminated. Each spatial arrangement of the sampling points has been optimized to maximize the rationality of the soil sampling layout which were based on the road network using simulated annealing algorithm. On this basis, the terrain factor and the soil organic mattter of the optimized samples were used to build multiple linear regression model, and establish a multilayer perceptron model based on neural network. The accuracy of the latter model and the precision of the former model were compared. The results show that: developing a soil sampling programe with the aid of using road network can be feasible, the optimized sampling point can be used to get accurate knowledge between soil and landscape, and better than the precision of the original sample. In this study, sampling designs were developed by using spatial distribution of roads, historical samples, digital elevation data and other available resources. This provides an efficient means and theoretical basis to reduce sampling costs, increase the sampling efficiency, and show the spatial distribution of the organic matter.(3) Taking the spatial distribution of soil organic matter in Zhongxiang City(search area A) for example, the terrain factors were generated from digital elevation model whose resolution is 30 m and integrated in the study, and classification rules of terrain units were built according to those of the combination for different levels of the terrain factors under each topographic condition. The area was exactly divided into 13 types of typical terrain units using the classification rules, and Ordinary Kriging was used for SOM interpolation and the spatial distribution of soil organic matter in corresponding area was accessing to based on the soil samples within different topographic unit. Those were done in order to fill the gaps of traditional land classification method which based solely on a single indicator(such as elevation). The spatial distribution of soil organic matters which contains inherent influence of topographic factors was obtained from the combination of the results in each terrain unit region. This study found that the similarity between the accuracy of the prediction result obtained from the more rugged terrain units and the accuracy of global prediction result reachs the degree of 0.75, while the accuracy of the prediction in the region of gentle terrain unit has significantly improved, the prediction accuracy improved by 16.39% than those of the overall prediction. so that the spatial characteristics of soil organic matter can be accurately and efficiently obtained by the spatial predicting method based on terrain units. Use geomorphology zoning to obtain a higher precision spatial distribution of soil organic matter, the synergistic effect of topography in the geostatistical study of soil organic matter was further explored. |
PDF Full Text Request