| Soils are highly variable spatially due to the combined effects of physical, chemical, and biological processes that operate with different intensities and at different scales. Consequently, uniform management of fields often results in over-application of inputs in areas with high nutrient levels and under-application in areas with low nutrient levels. Therefore, understanding of the spatial variability of soil properties is essential in determining local fertilizer needs of tobacco. Based on classical statistics, spatial variability of soil nutrients in a tobacco-planted field was studied and spatial distribution maps of soil nutrients was generated by the combined usage of GIS and geostatistics.The most popular approach to manage spatial variability within fields is the use of management zones (MZs), which are field subdivisions that have relatively homogeneous attributes in landscape and soil condition, and can be used to direct variable rate fertilizer application. To achieve these objectives, soil samples (0~20cm) were taken 81 points on an approximately 100-m grid in March 2007 using global positioning system to define sample locations. Soil chemical properties and texture were analyzed and their spatial variability was assessed by geostatistical techniques.Considering that most of soil properties are often related, principal component analysis (PCA) was used to summarize soil propertied into a few meaningful components (PC), which were further used to delineate management zones by fuzzy cluster algorithms. In addition, six field experiments were conducted to establish models for nitrogen, phosphorus and potassium fertilizer recommendation in Jia County, Pingdingshan, during 2007and 2008. This provides a basis of information for site-specific fertilizer management in tobacco-planted field. The main conclusions were as follows:(1) Coefficients of variation ranged from approximately 3.4% for pH to almost 48% for available Zn. Soil OM, TN, AN, AK, CEC and texture had medium CV (11.5%~23.4%). Soil AP and available microelement had higher CV, indicating the heterogeneity of soil properties. Thus, differentiated management may be necessary to achieve maximum economic and environmental benefits. Distributions of all the variables, except Fe, were only slightly skewed (skewness < 1), and their medians were close to their means. The Kolmogorov-Smirnov test revealed that all variables were normally distributed (P > 0.05) and did not require transformation. By log-transformation, soil available Fe showed normal distribution, thereby providing a basis for further structural analysis.(2) The results of semivariogram analyses are shown that coefficients of determination (R2) for all variables, except soil AN and available ZN, were greater than 0.87, indicating good fits. Soil pH, OM, TN, AP, AK,sand, clay and microelements were modeled best with spherical models, whereas CEC and silt with Gaussian and exponential models. Soil AN showed only pure-nugget effect fitted by linear model at sampling interval of 100-m, i.e., nearly horizontal linear semivariogram, indicating spatial independence. This suggested that a higher sampling density might be recommended for AN in this region. Ranges of spatial dependence varied from 274m (TN) to 1066m (CEC).(3) The ratio of nugget variance to sill variance can be used to classify spatial dependence of soil properties. A ratio<25% indicated strong spatial dependence, between 25% and75% indicated moderate spatial dependence, and >75% indicated weak spatial dependence. Soil properties with strong and moderate spatial dependence will be more readily managed and an accurate site-specific fertilization scheme for precision farming more easily developed. In this study, the geostatistical analysis suggested that most soil properties showed moderate to strong spatial dependence. Soil pH, TN, Fe, Cu, Zn, sand and clay were strongly spatially dependent with nugget/sill ratios between 16% and 24%, respectively. Soil OM, AP, AK and CEC were moderately spatially dependent with nugget/sill ratios ranging from 30% to 50%. Soil AN showed no spatial dependence. Soil AP and AK had significant correlation with intrinsic properties, for example, pH, OM, and clay, indicating the low extrinsic component of variability and could be used as the basis for differentiated fertilizer application in this region.(4) PCA and fuzzy cluster algorithm were then performed to delineate MZs. The first two PCs with eigenvalues greater than 1 were considered for clustering analysis, which was performed in MZA procedure using fuzzy c-means cluster algorithm. Performance index (FPI) and normalized classification entropy (NCE) were used to determine the optimum cluster number. Results showed that the optimum number of MZs for this study area was three and analysis of variance indicated the heterogeneity of soil fertility among different MZs, which also showed the combined usage of PCA and fuzzy cluster algorithm is an effective method to characterize spatial variability of soil properties within MZs. The defined MZs provide a basis of information for site-specific fertilizer management in the tobacco-planted field.(5) By determining the parameters correlative with Truog-Stanford equation building, the models for nitrogen, phosphorus and potassium fertilizers recommendation were established based on two years field experiments. Using the three equations, the fertilizer application for three management zones were recommended respectively.The innovation points for this study are as follows:(1) Using GIS and geostatistical techniques, spatial variability was assessed and contour maps generated for soil nutrients in a tobacco field. Establishing a procedure to delineate management zones with the combined usage of PCA and fuzzy cluster algorithm, this provides a basis for defining management zones for tobacco field.(2) Determining correlative model parameters and initially establishing fertilizers recommendation equations of nitrogen, phosphorus and potassium for flue-cured tobacco in this region. |
