Study On Estimation Of Total Salt Content And PH Value Of Saline Soil In Tuoketuo County Based On Hyperspectral

This paper took the saline soil of Tuoketuo County as the research object,and the measured soil hyperspectrum,total salt content and pH value as the data source.A variety of mathematical transformations were performed on the hyperspectrum,to select the sensitive bands,calculate the spectral index,and an inversion model of soil total salinity and pH value based on sensitive band and spectral index was established.An evaluation based on model indicators was performed.The main research conclusions are as follows:(1)The salinity in the study area was serious.The maximum total salt amount was 84.90g/kg,and the minimum value was 16.10g/kg.The mean and coefficient of variation were large,and the mean was greater than the median.The data was positive,and the amount of more than half of the soil total salt was greater than 34.60g/kg,the total salt content of soil samples was mostly 35-40g/kg;the pH value was between 9.28-10.52,and the median value was less than the mean value,indicating that the pH value of most soil samples was less than 9.97,and the study area as a whole had strong alkalinity,which was more difficult for crop growth.(2)In this study,the spectral curve of saline soil belonged to the "steep ridge type".The spectral reflectance generally increased first,then gradually flattened,and then decreased.As the total salt content and pH value increased,the spectral reflectance also increased.The slope of the indoor spectrum was greater than that of the outdoors,and the absorption valley was more pronounced at 1400nm and 1900nm outdoors.The slope of the spectrum in the visible range was:green light>blue light>red light>near infrared.The local maximum extraction method was used to find the spectral curve.There were 20 reflection peaks.After continuum removal treatment,there were 8 obvious absorption valleys,most obvious at 1900nm.(3)The correlation between indoor spectrum and total salt content and pH value was greater than outdoor,and the correlation between indoor spectral reflectance and total salt content from 11 mathematical transformations in order from large to small was:inverse first order differential>log first order differential>first order differential of reciprocal logarithm>arithmetic square root first order differential>first order differential>untransformed reflectance>inverse number>logarithmic>arithmetic square root>logarithm of reciprocal>second order differential>normalization,pH correlation from big to small,the order was:first order differential>second order differential>arithmetic square root first order differential>log first order differential>reciprocal first order differential>reciprocal first order differential>untransformed reflectance>arithmetic square root>pair number>reciprocal log>reciprocal>normalization.The correlation of the first-order differential transformation was higher than other processing methods,so the first-order differential transformation was a better processing method when extracting spectral information.(4)The accuracy of the least squares support vector regression model based on sensitive bands was better than the multivariate stepwise regression model,and the R2 value of the model was above 0.9.The accuracy of the single salt index model of total salt was from high to low:index RI model>difference soil index DI model>salinity index 3 SI3 model>salinity index 2 SI2 model,R 2 value of difference soil index RI model was 0.624.The order of the accuracy of the single spectral index model of pH value from big to small is:salinity index 2 SI2 model>salinity index 3 SI3 model>salinity index S3 model>difference soil index DI model>difference soil index RI model,the R2 value of Salinity index 2SI2 model is 0.667.The accuracy of the multiple stepwise regression model based on spectral reflectance,single index,and composite index was:index comprehensive model>single index model>spectral reflectance model.Among them,the least square support vector regression model and the comprehensive index model have high precision,which can predict the total salinity and pH value of saline soil in this area.