| Soil nutrient were related to the quantity and quality of crops. Fertilizers played a key role to replenish soil nutrients. According to the soil nutrient situation, an appropriate fertilizer can be developed to add the appropriate nutrients, conduct restoration of land and increase the vitality of the land. Therefore, the study of the distribution of soil nutrients for crop growth is really an indispensable important and necessary part of the development of agriculture and forestry.In this paper, the soil of one Gannan Orange orchard was taken for the study. As total nitrogen, organic matter and total phosphorus nutrient information on different layer of soil were really different, Fourier near-infrared spectroscopy combined with visible / near-infrared spectra of different parameter detection were involved to collect soil sample spectrum. First of all, measured nutrient value was determined by chemical analysis combined with spectral pretreatment methods and chemometric algorithms to establish quantitative model. The optimal choice of instrument parameters, the most suitable pretreatment methods and algorithms were set up to determine the final mathematical model. Then, the nutrient content of the predicted value and the measured value of nutrients in the optimal mathematical model combined with statistical analysis methods carried on nutrients interpolation to get the true value of soil nutrients and spatial distribution map of prediction set in the study area under the profile. Finally, the national standard classification of soil nutrients to be used to formulated soil nutrients value interpolation within the reasonable area. Variable soil fertilizer prescription map was formed then to guide the precise fertilization.1. The experimental equipment and parameters used were described. The near-infrared Fourier Spectrometer(TENSOR37) and visible / near-infrared detector instrument parameters was made optimize contrast. Partial Least Squares combined with full cross-validation method to get best match parameters of soil samples for each detection instruments. At the same time, the paper introduced the experiment in physical and chemical analysis of soil nutrient testing methods and formulas. Soil nutrient sampling methods, sampling point distribution and spatial distribution analysis software also be involved.2. The spectral of soil samples for each different layer were processed with different preprocessing analysis as S.G. smooth, standard normalization(SNV), multiplicative scatter correction(MSC), baseline correction, the first derivative(1st D), second order differential(2st D). By partial most squares(PLS), principal component analysis(PCR), the least square-support vector machine(LS-SVM) and other chemometric modeling methods, the appropriate quantitative and comparative analysis of models were established. The results showed that pretreatment done slightly different for different layers of their nutrient. But the indicators of different soil layers are best using a effective model of PLS algorithm. The correlation coefficient of topsoil soil total nitrogen, total phosphorus and organic matter determination with RP were 0.833, 0.718 and 0.714 and RMSEP were 7.622, 0.098 and 1.515, respectively.The correlation coefficient of heart soil total nitrogen, total phosphorus and organic matter determination with RP were 0.809, 0.762 and 0.743 and RMSEP were 4.473, 0.107 and 1.271,respectively. The correlation coefficient of substrate soil total nitrogen, total phosphorus and organic matter determination with RP were 0.805, 0.806 and 0.794 and RMSEP were 3.359,0.052 and 1.388, respectively. The test also discussed the relationship of various layered soil nutrients and plant growth. The correlation analysis showed that the distribution of nutrients are closest in the subsoil and plant growth conditions. The other two soil layer also had some reference value for nutrient content analysis determined by the relationship between the growth of the plants and soil nutrient content.3. The number of sampling points is limited in this experiment. In order to obtain the distribution of nutrient data for the entire study area, the use of Kriging(Kriging)interpolation of these sample points are required to conduct interpolated for nutrient of each point. Paper used module of Arc GIS Geostatistical analysis software to operate interpolating operation for sampling points of total nitrogen, total phosphorus and organic matter to get the real value of the chemical and the optimal predicted model values. The draw soil nutrient spatial distribution map of the sample area were obtained, with the state of soil nutrient classification table, the fertilizer required administration also carried out the optimal Kriging interpolation and obtained a more variable soil nutrient fertilizer prescription map.In summary, spectra collected by FT-NIR spectrometer, applying pretreatment combined PLS algorithm established the best model in the experiment. The subsoil and plant nutrient distribution were closest to the actual growth conditions through the research. |
PDF Full Text Request