Study On Remote Sensing Monitoring Of Soil Salinization Based On Spectral Information

Soil salinization is the main ecological problem that restricts vegetation growth in arid area.It is also one of the main obstacles to land use and agricultural development.The salty land is considered to be an important reserve cultivated land resource in China,the timely and accurate acquisition of soil salinization information is very important to control the saline soil,prevent its further degradation and promote the sustainable development of agriculture.The problem of soil salinization is very prominent in the Northern Yinchuan plain,and the secondary salinization caused by irrational use of pesticide and chemical fertilizer aggravates the degree of soil salinization.Therefore,it studies the effect of soil salinization in the north of the Yinchuan area,and obtains the dynamic change information of soil salinization in real time and accurately,which can prevent the salinization from worsening and protect the sustainable development of the ecological environment.Remote sensing technology can acquire surface information in large area and timely and accurately,and is a powerful and effective technology for quantitative monitoring soil salinization.This project selects typical soil salinized area,Pingluo County in the northern plain of Yinchuan as the research area.Using remote sensing images and field measured spectra as data sources,the sp Northern Yinchuan,Ningxia ectral characteristics of saline soil and typical vegetation are systematically studied.The best spectral transformation methods and response bands related to soil salinization are selected from the measured spectral data of saline soil.The prediction model of soil salinization information is built,and the similarities and differences between the field measured spectral characteristics and the remote sensing image spectra are determined,and the spectral information of the remote sensing images is corrected by using the field spectrum.Finally,the prediction model of soil salinization information based on remote sensing images is established to realize the large area,real-time,and real-time information of the saline alkali land in the same area.The scientific basis for accurate prediction is provided.The main conclusions of this study are as follows:(1)The overall change of soil spectral curve is gentle and has certain volatility,the reflectance of soil surface had no obvious peak and valley,the positive correlation between bands is good.On the whole,the higher reflectance,the higher corresponding peak.The partial reflectance of visible light is small.It has increased in the near infrared band,but not much.The soil spectral curve of the study area accords with the basic characteristics of the four types of the middle slope curve.In the visible light band,the spectral curves of different salinized soils do not show strict regularity.The overall change of vegetation's spectral curve is large,the fluctuation is strong,with obvious peaks and troughs,and the correlation between the bands is poor.On the whole,the spectral curves of different salinization degrees are basically similar,the trend is different with the increase of wavelength.Between 770 nm and 1300 nm,with the increase of soil salinization,vegetation reflectance decreased gradually.(2)The coif4 wavelet is selected and the decomposition scale is 2 layers,the threshold is heursure and the threshold adjustment is sln,the de-noising effect is best for the hyperspectral data of salinization.The experiment shows that the wavelet threshold processing is an effective method of spectral denoising,which is beneficial to the establishment of a hyperspectral soil salt content inversion model with more stable and higher prediction accuracy.Therefore,wavelet denoising can be applied to the hyperspectral monitoring model,and it is widely used for remote sensing estimation of large salty soil.(3)The transformation effect of cosine differential in vegetation spectrum is the best,and 686 nm and 1461 nm of vegetation spectrum were selected as the best sensitive bands of soil salinity.Establishment of hyperspectral prediction model based on soil adjusted vegetation index MSAVI,the determination coefficient R2 reached 0.77,through the 0.01 significant level test,this model can be used to predict soil salinity in arid area.Soil spectrum 416 nm,543 nm and 953 nm wavelength were selected as sensitive bands of soil salinity.There are higher correlation between si2 salt index and soil salinity,correlation coefficient is greater than 0.6.Using polynomial regression algorithm,the best estimation model of soil salt content based on soil spectrum can be established.(4)The correlation between salt index si1 and total salt content is the best,and using si1 to establish a prediction model of salt content,the determination coefficient R2=0.68,through the 0.01 level significant test.Therefore,it can provide a basis for fast and accurate prediction of salinization information of large area of cracked soil.