Spatial Variation and Site-specific Management Zones of Soil Nutrition in Tobacco-planting Fields of Luzhou Based on GIS and Geostatistics
|School||Henan Agricultural University|
|Keywords||flue-cured tobacco tobacco-planting fields GIS spatial variation management zones|
Soil, whose fertility conditions impacts the quality of flue-cured tobacco, is the basis for tobacco production and a continuous spatiotemporal variant. Soil is characterized by high heterogeneity, which hasn’t been taken into account in our traditional farming. Fertilizing according to experiences for one thing is lack of direction and could easily result in atmosphere and water pollution and low utilization rate of fertilizer, for another it couldn’t enhance the tobacco quality and peasants’income, either. Therefore, it is highly significant to understand the spatial variability of soil nutrients, and then divide management zones and practice variable rate fertilization. Modern tobacco agriculture calls for variable rate fertilization, enhancement of fertilizer utilization rate, and at the same time protecting the ecological environment.Soil samples (0~20cm) were taken on an approximately 20-m grid with GPS technology in the experimental areas of tobacco research institute of Luzhou, Sichuan province, and by the combined usage of GIS and geostatistics, this paper analyzed the spatial distribution pattern and provided the distribution map for the soil nutrients within the research area. Taking the principal components as data source, the management zones were divided through fuzzy c-means cluster algorithm, which provided basis for the site-specific management of soil nutrients in this area. Results were as follows:(1). the coefficients of variation of soil nutrients were between 14.24 % and 72.05%, all of which were medium. That meant that the spatial heterogeneity was relatively higher and it was necessary to adopt differentiated management for the soil nutrients. Soil pH had the lowest coefficient of variation, namely 14.24 %, while soil copper had the highest one, namely 72.05%. For major element in soil, such as nitrogen, phosphorus, potassium, the coefficients of variation were between 27.45% and 44.95% and relatively lower, while for microelement, like iron, manganese, copper and zinc, they were between 45.60% and 72.05 % and relatively higher. This probably was related to the actual production which paid more attention to the application of major elements while ignored microelements, thus it was strongly suggested that microelements fertilizer should be applied properly in the future. The skewness of all items were from 0.051 to 1.61 and all the skewness statistics were positive values, which indicated that all the distribution curve sloped to the left. The kurtosis of all soil nutrient indices were from -1.07 to 6.83, kurtosis values of all the items except pH, OM, AP, Mn and Zn were positive, meaning a peak distribution.(2). the classical statistics of soil properties could only describe the total changing situation, but could not reflect the spatial variability of soil properties quantitatively. In this paper the spatial variability of soil properties in this study area was probed through geostatistics. The Kolmogorov-Smirnov (K-S) inspection demonstrated that all soil properties were normally distributed will Pk-s>0.05 except copper, by log-transformation, soil copper showed normal distribution. The analysis of geostatistics showed that the theoretical model for AK was linear with higher residual error and lower degree of fitting, which revealed pure-nugget effect. The optimal theoretical model for AN, AP, TP, Fe and Mn was exponential model, and spherical model was best for other soil properties. The ratio of C0/(C+C0) for TP, Fe, AP, Mn and AN was 66.7% to 72.9%, which was medium spatial correlation and demonstrated that their variation was caused by combined effects of structural and chaotic factors. Under the sampling scale in this paper, AK changed smoothly and was completely chaotic variation; its spatial correlation was less. The ratio of C0/(C+C0) for other soil properties were 76.2% to 99.8%, which indicated the spatial correlation was relatively weak. The biggest correlation distance (range) for different soil nutrients in this study area varied sharply and were from 24m to 85.31m, which meant that factors affecting soil properties functioned under different scale.(3). in this area, each soil properties showed apparent spatial distribution. The pH value increased from northeast to southwest and the lower values were in north and northeast parts while the higher ones were in the middle and south parts. The distribution of OM presented as strip and plaque and increased from southwest to northeast, the lower values were mainly in the middle and southwest parts while the higher values were in the north and northeast. The values of AN in the north parts were lower and in the south parts were higher, and the lowest values were in the north while highest values were in the south, contour maps of different values showed as narrow strips. The distribution situation of TP and AP were almost the same, the lower values were in middle and northwest while the higher values were mainly in the middle and southeast and were often divided by lower valued plaques, indicating that random factors like anthropogenic in field influenced the distribution of phosphorous heavily. The distribution of soil CEC was typical plaque, with higher and lower alternation. The distribution characters of TK and AK were similar, the levels of both were high, especially TK the lower parts of which was odds and ends. The lower parts of AK were limited in a small area in the northeast. It was advised that these study areas should be applied iron, manganese, copper, zinc fertilizer, as the content of these microelements were lower generally.(4). The soil nutrients management zones were defined through fuzzy c-means cluster algorithm and principal component analysis, and the optimal numbers of management zones were decided by the indices of performance index (FPI) and normalized classification entropy (NCE). The sum of the four principal components in the paper could explain 60% of the total variation, and the Eigen value of each principal component was >1. The score of the four principal components were introduced to the MZA software to cluster. Both NCE and FPI reached their lowest values when the zone number was three, accordingly, the optimal theoretical management zones was three. The fertilization model and volume recommended were determined for every management zone, which laid a foundation for site-specific fertilizer management of the tobacco-planting soil in this research area.