Research On Quantitative Inversion Of Soil Salinity With Multi-source Spectrum Data In Arid Area

Soil salinization is one of the ecological environment degradation problems in the world,especially in the arid and seme-arid regions.Serious soil salinization led to degradation of vegetation and desertification,and causes negative effect on the ecological environment and agricultural sustainable development.Soil salinization is the most prominent problem in Yutian oasis,Xinjiang,China,it threaten to the local agricultural production and the development of social economy.Using the measured data and remote sensing technology can improve the monitoring accuracy of soil salinization,and it can provide basic theores and data support for the real-time monitoring of soil salinization.This research analysis the soil physical and chemical properties and soil reflectance spectral characteristics with laboratory reflectance spectrum,field reflectance spectrum,and WorldView-2 multi-spectral reflectance spectrum.Analysis the correlation between soil spectral reflectance and soil salt content,water content,pH value,and electrical conductivity.According to the analysis results,combining with the basic spectral characteristics of the saline soil,select the sensitive wavelengths of soil salinity.Using partial least squares regression and stepwise multivariate linear regression analysis method prediction modle of soil salinity,soil moisture,pH value,electrical conductivity with the measured spectral data.Using partial least squares regression and stepwise multivariate linear regression analysis method prediction modle of soil salinity,soil moisture,pH value,electrical conductivity with WorldView-2 multi-spectal reflectance spectral data.Using the determine coefficient(R~2),modeling total root mean square error(RMSE),the residual prediction deviation(RPD)evaluate the accuracy of model.Mapping the spatial distribution of soil salinity,soil moisture,pH value,electrical conductivity.The main contents and conclusions are as follows:(1)The soil salinization degree analysis results show that the soil salt content,water content,pH value and conductivity content in the study area is generally higher,difference is bigger,and the soil pH range is 7.61-9.86.The spatial distribution of soil salt parameters is not uniform,and the tendency of a certain gathering is present in the distribution of space.(2)Laboratory reflectance spectra,field reflectance spectra and WorldView-2 multi-spectral image were analyzed to get the basic saline soil spectra,and the correlation between soil salinity,soil moisture,pH value,electrical conductivity and spectral reflectance were analyzed.The results show that laboratory reflectance of the soil samples in 400-600 nm band has risen sharply,from 600 to 800 nm rising slowly,while in 800-1900 nm band reflectance rise more slowly,in 2150 nm reached maximum value,then began to decline.WorldView-2 multispectral image reflectance in visible a sharp rise in the 450-625 nm band spectral reflectance,630-745 nm band reflectance slowly rising,while 745-1040 nm band reflectance more slowly.The saline soil of different salinity laboratory and field spectral reflectance spectrum curve reflects the 600 nm,800 nm,1000 nm,1400 nm,1900 nm,2100 nm,2200 nm,2400 nm near absorption peak with the increase of salinity has deepened.(3)Spectral reflectance of the first order differential and second order differential is the best form of transformation,the soil salt inversion it can make the small spectral characteristics of the amplifier,the spectral reflection curve of easy to observe,the small spikes appears at 1000 nm,1400 nm and 1800 nm,and at 1400 nm,1950 nm,1950 nm appears more apparent spike,suggests that these bands can effectively reflect the characteristics of soil salt.Build the prediction model of soil salinity,soil moisture,pH value,electrical conductivity with PLSR and SMLR method based on the measured spectra and WorldView-2 multi-spectrum image reflectance spectra.Correlation analysis results show that the laboratory spectral reflectance of first order differential and the correlation coefficient of soil salinity peaked at 1950 nm,field measured spectral reflectance of first order differential and the correlation coefficient of soil salinity peaked in the 795-1000 nm band range.In laboratory spectral reflectance of second order differential and the correlation coefficient of soil salt content,450-1000 nm wavelength range is the main area of spectral response of soil salinity,the measured spectral reflectance in the field of second order differential and the correlation coefficient of soil salt content,800-1350 nm wavelength range is the main area of spectral response of soil salinity.The correlation coefficient of WorldView-2 multi-spectral reflectance spectrum and soil salinity was peaked at 705-745 nm,770-895 nm,and 860-1040 nm.(4)PLSR and SMLR method respectively based on the laboratory and field survey the soil salt content of the spectrum inversion model: the stability of the integrated model as well as the ability to predict,the soil electrical conductivity is the best salt parameter inversion of soil salt content,based on field measured spectral data of partial least-squares regression model is the best inversion model of soil salt content and R~2 is 0.75,RMSE is 0.05,RPD is 1.46.(5)The correlation between laboratory,field and the World View-2 multi-spectrum reflectance spectra is best,the field and indoor measured spectral reflectance shows close relationship,the high value of correlation coefficient is 0.99.The trend of measured reflectance spectra and WorldView-2 multi-spectral reflectance spactra was very similar.(6)WorldView-2 multispectral images of the three soil salt sensitive wave bands,NDVI,SI and other five parameters as the input parameters of the model established based on PLSR and SMLR inversion model of multispectral images,the results showed that soil salinity high value area mainly concentrated in southwestern and northern areas of the river.Based on a WorldView-2 multispectral images of the partial least-squares regression model is the best inversion model of soil salt content and model of the determination coefficient R~2 is 0.78,RMSE root mean square error is 0.06,the residual prediction deviation RPD is 3.04,the soil electrical conductivity is the best salt parameter inversion of soil salinity.