| Agricultural production is the basis for the development of modernization in China, and it is related to national economic development and social stability. Soil is fundamental to agriculture, so understanding the changing conditions of soil is the basis for the sustainable development of agricultural production. Soil is a synthesis with extremely complex pattern and evolution process. Soil properties include spatial variability and spatial dependence. These properties have critical effects on improving cutting-edge development of many disciplines, such as agricultural soil, agricultural irrigation, precision agriculture et al. Since the method of spatial analysis was introduced, the research on spatial variability and dependence has been an academic focus.Zhejiang province has rich tideland resources, and since the founding of China, 240 thousands of acres of tideland has been built, which has become a major commodity production base for grain, cotton, aquatic products and fruits, and significantly eased the contradiction between less and more. Because soil of tideland mainly results from sediments in river and sea, soil there has many problems, such as heavy saline-alkali and poor fertility. Besides, tideland reclamation is short of freshwater and in typhoon areas, resulting in low crop yield, low soil productivity, and low economic benefit. In order to promote tideland from traditional agriculture to modern agriculture, increase the efficiency of nutrient utilization, reduce the amount of fertilizer, decrease production cost and increase yield, it is necessary to understand the amount and transformation of soil nutrients, clarify the effects of soil nutrients to cotton yield and adopt scientific management zoning.Viewing the obvious limitation of classic statistical analysis in spatial geographic data or phenomena, spatial analysis was applied, which considers spatial variability and spatial dependence. Through Moran's I and LM-test, the robust of spatial lag model and spatial error model were compared. Spatial lag model was selected to complete regression of soil nutrients and cotton yield in study area. Using 7 and 4 soil characteristics, the fitting degrees of classic regression models were 44.24% and 40.54%, and the corresponding fitting degrees of spatial regression model were 55.33% and 51.99%. It was shown that spatial regression model had higher fitting degrees. Kriging interpolation maps for residuals were drawn. The residuals of classic regression and spatial regression had high consistency, and the distributions were in line with the soil nutrients.Readily available k, organic matter, cation exchange capacity and total nitrogen were selected due to their greater impact on crop yield. EM38 conductivity instrument was introduced to gain 526 samples of soil electrical conductivity. Principal component analysis was made through Kriging maps of sampling data, and the first two principal components extracting 99.99% cumulative contribution rate were picked up. Management zoning was conducted by fuzzy c-means clustering. Fuzzy performance index (FPI) and normalized classification entropy (NCE) were provided to evaluate the best number of zones. Besides, fuzzy k-means classification was carried out. The results of the two classifications had a high degree of consistency, and the best number of zones is 2. Finally, zoning results were compared and validated by analysis of means. It was shown the management zones had good space continuity. Besides, the soil nutrients and cotton yield had significant differences in sub-districts. The management zoning could guide the precise fertilizer and production management, and improve the crop production of tideland reclamation.
|
PDF Full Text Request