| With the increasing needs for natural resources, human beings are changing the land system in pervasive ways, which leads to the degradation of environmental conditions. To reveal the impacts of the human disturbances on land quality, some international organizations, e.g., FAO and UNEP and many governments, have committed global and regional land quality investigating, monitoring and evaluating works. China, as a thickly populated country with scarce land resources, faces the dilemma between economic development and land resources preservation, and needs to investigate land quality precisely for making scientific policy of land management.During the investigations, spatial sampling method is one of the major data acquisition techniques, which is capable of reducing labor, material and financial cost on the premise of assuring high precision. However, traditional sampling methods cound not discrimate the spatial patterns of land quality indicators, and suffered from low sampling precision and efficiency either. Therefore, it is necessary to develop intelligentialized and efficient spatial sampling technique for land quality monitoring, which takes the spatial patterns of land quality indicators into account.Particle swarm optimization (PSO), a typical population intelligence optimization (PIO) algorithm, can simulate the cooperation and competition among individuals and address the complicated optimization issues. Due to its fast convergence rate, intelligence and compatibility, the PSO has been applied in many fields, e.g. vehicle routing, shop scheduling and land use allocation issues. Accordingly, this dissertation proposed a series of optimal sampling models for land quality monitoring on the basis of the PSO algorithms with different optimization objectives and constraints, including a single-variable soil sampling model, a multi-variable cooperative soil sampling model and a spatial sampling model for urban land quality monitoring. These models considered the differences between the spatial patterns of natural land quality indicators and of social and economic ones, and attempted to optimize sample size and locations simultaneously.The major works and contributions of this dissertation are as follows:(1) Two spatial sampling schemes, including geostatistic-based rural land sampling scheme and spatial clustering-based urban land sampling scheme, are systematically summarized based on the principles of land quality and spatial sampling. In each scheme, the sampling scenarios are designed according to the spatial patterns of land quality indicators. Furthermore, the principle and the framework of Particle Swarm Optimization algorithm (PSO) are described in detail and its potential applications for sampling design are presented.(2) A new optimal sampling model for single soil variable namely Integrated Sampling method incorporated with Binary Particle Swarm Optimizer (BPSO_SV) is proposed based on geostatistical sampling scheme (which is also called model based sampling) and design based sampling theory. Considering minimum kriging variance (MKV) and maximum entropy (ME) as the optimal objectives, and minimum sample capacity, spatial sampling barriers and sampling accessibility as the constraints, the model is capable of optimizing sample size and locations simultaneously. In addition, the BPSO_SV model employs voronoi and semivariogram analysis techniques to stratify and partition the sample population to guarantee the regional characters of sample units and stratifications. Compared with traditional sampling methods in the test, the BPSO_SV model obtains the optimal sampling solutions with lower MKV values and higher ME values, and features faster convergence rate.(3) A novel sampling model namely multiple soil variables cooperative sampling model incorporated with cooperative particle swarm optimization algorithm (CPSO_MV) is proposed based on geostatistical sample scheme. The model organizes the individuals and sub-swarms with ring structure, and takes minimum kriging variances of each soil variable as the optimal objectives and minimum cooperative sample size, sampling accessibility and sampling cost as the constraints. Additionally, the CPSO_MV utilizes variable precision rough set (VPRS) to select the key land quality indicators (LQIs), and quantitatively analyzes the spatial relationships between LQIs. The experimental results demonstrate that the CPSO_MV can reconcile multiple soil variables better than the traditional methods, e.g., weighted aggregation (WA).(4) According to the spatial distribution patterns of urban land quality indicators, a novel sample model namely adaptive spatial cluster sampling method incorporated with particle swarm optimization (ACSPSO) is proposed based on spatial cluster sampling scheme. The model takes maximum distances between classes as the optimal objective to generate original sample points with the application of the PSO clustering algorithm on the basis of design based sampling theory, and then considers maximum entropy and WM criterion as the objectives to set extra points around the original ones on the basis of model based sampling theory. The optimization of the sampling solution is restricted by accessibility and distribution centers of sampling points. The experimental results show that the ASCPSO can accurately describe the clustering distribution pattern of sampling points, and decline the WM value by 29.18% and promote the ME value by 3.83% compared with the traditional geostatistical-based sampling method (GSPSO).(5) Spatial Sampling Decision Support System for Land Investigation (SSDSS-LI) is designed and completed based on the theory of Spatial Decision Support System (SDSS) and the optimal sampling design techniques proposed in this dissertation. SSDSS-LI system consists of three levels and has the spatial sampling functionalities oriented rural and urban land quality monitoring. Experimental results have proved the feasibility, validation and extendibility of the system.
|
PDF Full Text Request