| Taking Tuoketuo County of Hohhot City as the research area,this thesis selected typical plots of different land use types and different soil types in the research area,collected soil samples on the spot,determined the contents of organic matter,total nitrogen,total phosphorus and total potassium in the laboratory,and used SVC Hr1024spectrometer was used to determine the spectral reflectance of soil samples.Through mathematical transformation and sensitive band selection,the best spectral inversion model of soil nutrients was established.The research results can provide a theoretical basis for soil nutrient monitoring in the study area and even tumechuan plain,and provide a certain technical support for ecological environment construction.The main conclusions are as follows:(1)Soil nutrients and their distribution in the study area:There was a positive correlation between the contents of soil organic matter,total nitrogen,total phosphorus and total potassium;the spatial distribution was high in Northwest and low in southeast,and the overall gap was large;the average content of soil organic matter in Tuoketuo county was in an urgent shortage level,the average content of total nitrogen and total potassium was in a lack level,and the average content of total phosphorus was in the middle and lower level;under different land use types,the average distribution of soil nutrient was cultivated land>woodland>grass The average distribution of soil nutrients under different soil types was chestnut soil>sandy loam soil>saline alkali soil.(2)The difference of Soil Spectrum and the improvement of correlation coefficient:The correlation coefficients of soil organic matter,total nitrogen,total phosphorus and total potassium increased by 0.20,0.14,0.11 and 0.16 respectively after continuous wavelet transform,and the prediction ability of the model was improved as a whole.Continuous wavelet transform can improve the correlation between soil nutrients and spectrum The correlation coefficient of the model is of great significance to the prediction ability of the model.It is of great significance for the prediction of soil nutrients in this region.(3)Optimal model selection:the prediction model of support vector machine,the prediction effect of RBF kernel function is higher than that of linear kernel function,polynomial kernel function and sigmoid kernel function,but there is instability in the model,and the environment setting of the model is also too high,so support vector machine can not realize the prediction of total phosphorus and total potassium content in the study area;the parameters of BP neural network prediction model are relatively simple,and different prediction data are selected Different number of neurons can effectively predict soil nutrients in the study area,and the prediction effect is higher than that of support vector machine.(4)The best prediction model results of different soil nutrients:Bp-CWT-R was the best prediction model of soil nutrients in this study,The suitable soil types are mainly sandy loam,chestnut soil and saline alkali soil,and the land use types are mainly cultivated land,forest grassland and saline alkali soil in Tuoketuo County of tumechuan plain;R~2=0.76,RMSE=2.14,R~2=0.60,RMSE=1.83,RPD=2.02,R~2=0.77,RMSE=0.19,R~2=0.70,RMSE=0.15,RPD=2.40,R~2=0.71,RMSE=0.12,R~2=0.61,RMSE=0.07,RPD=2.40,R~2=0.71,RMSE=0.12,R~2=0.61,RMSE=0.07,RPD=0.02,The best prediction model of total potassium is R~2=0.65,RMSE=0.52,prediction set R~2=0.58,RMSE=0.44,RPD=2.09,and the RPD of the model is above 2.00. |
