| This paper takes the urban traffic green space soil as the research object,analyzes the spatial variability and distribution characteristics of soil magnetic parameters in traffic green space,and supports the correlation between the magnetic parameters and different soil hyperspectral transformation forms with the support of urban traffic green space soil hyperspectral data.To establish a hyperspectral inversion model of soil magnetic parameters and verify it,the main conclusions are as follows:?1?The statistical results of the soil environmental magnetic parameters in the traffic green land show that the magnetic minerals in the soil are mainly ferrimagnetic minerals and contain a small amount of incomplete antiferromagnetic minerals.The magnetic mineral grains in the traffic greens are pseudo-domains?PSD?.Domain?MD?dominates,with less superparamagnetic particles.?lf,?fd%,?ARM,SIRM and HIRM are moderate variations,and S-100mT100mT and S-300mT300mT are weak mutations.?2?The fitting results of the semi-variance function model show that the soil?lf and SIRM accord with Gaussian model,the soil?fd%,?ARM,HIRM and S-300mT300mT accord with the spherical model,and the soil S-100mT100mT conforms to the linear model.Soil?lf,?ARM,SIRM,HIRM,and S-300mT300mT have a strong spatial autocorrelation and are mainly affected by structural factors.The spatial autocorrelation of?fd%is weak.?3?The spatial distribution map and error map of magnetic parameters in soil surface were obtained by ordinary Kriging optimal interpolation method.In general,?fd%was distributed in multiple islands and S-100mT100mT was distributed in bands.The remaining soil magnetic parameters were generally characterized by the spatial distribution of bands and islands.Soil magnetic parameters(except for?fd%and S-100mT)of traffic roads showed a trend of increasing first and then decreasing with the increase of vertical distance of the road,which was expressed as a“low-high-low”distribution pattern,which was associated with smoke emitted by vehicles.The particle distribution is related to the sedimentation process.?4?The spectral reflectance curve of the soil environmental magnetic parameters of traffic green space showed an increasing trend,and the degree of dispersion was relatively large.The shape was steep at 400600nm,and the slope was relatively large.This was related to the iron contained in the soil.The overall trend at 6001400 nm was overall.And through spectral transformation can effectively improve the correlation between soil magnetic parameters and spectral reflectance of soil,and is conducive to the establishment of models based on spectral reflectance soil magnetic parameters content.?5?Based on the magnetic environmental parameters of the traffic green space soil,a hyperspectral inversion model for six soil magnetic parameters was established using linear?multiple stepwise regression,partial least squares regression?and nonlinear?artificial neural network?methods,respectively..The verification results show that the BP artificial neural network model is superior to the low-frequency magnetic susceptibility?lf,non-magnetic remanent magnetic susceptibility?ARM,hard remanence HIRM,and demagnetization coefficient S-300mT300mT modeling sets and verification sets.Multivariate stepwise regression model and partial least squares regression model.When modeling the saturated isothermal remanence SIRM,the multivariate stepwise regression model is more accurate than the other two models.When modeling the demagnetization coefficient S-100mT,the modeling and prediction accuracy of the partial least-squares regression model is higher than that of the other two models.Through the comparison of the six magnetic parameters,it can be found that when modeling the low-frequency susceptibility?lf,,the model has the best stability and accuracy,and the best model is the BP artificial neural network model. |
PDF Full Text Request