Characterization And Hyperspectral Quantitative Model Of Iron Oxide In Red And Yellow Soils Of Mountainous Areas

Iron oxide is the main body of iron-bearing minerals in soil,and its accumulation and migration activities can reflect soil-forming process and soil-forming environment,which is one of the indicators of soil development and soil classification.The traditional method of soil iron oxide determination has problems such as complicated process,long cycle time,high cost and difficulty in obtaining data in time.With the booming development of hyperspectral remote sensing technology,rapid,economical and non-destructive acquisition of soil component information can be achieved,providing a new scientific way for the identification and quantification of soil iron oxides,which is of great significance for understanding soil properties and promoting numerical soil classification.Soil samples of mountain red loam,mountain yellow-red loam,mountain yellow loam and mountain yellow-brown loam developed from granite were collected in Dawei Mountain,Liuyang City,Hunan Province,as the study area,and the contents of soil total iron,free iron,amorphous iron and major iron oxide minerals(goethite and hematite)were determined.Based on the hyperspectral data of soils collected by soc710 imaging hyperspectral spectrometer,the spectral curve characteristics and characteristic bands of soil total iron and main iron oxide minerals(goethite and hematite)were determined by Savitzky-Golay convolution smoothing,reciprocal transformation,logarithmic transformation,square root transformation,differential transformation,continuum removal and correlation coefficient method.The quantitative hyperspectral inversion models of iron oxide in mountainous areas of specific regions were constructed based on three modeling methods:unitary regression,multiple stepwise regression and partial least squares regression,respectively.The main findings are as follows:(1)The characteristics of the vertical evolution of soil iron oxide morphology and contentSoil samples from the study area had iron properties,with an increase in soil organic matter content contributing to the enrichment of amorphous iron and an increase in soil acidity promoting the formation of free iron.With increasing altitude,hydrothermal conditions gradually decreased,soil development decreased,and the content of total and free iron oxide decreased significantly(P<0.01),and the content of amorphous iron increased significantly(P<0.01);the free iron content decreased and the iron activation increased,and both reached significant levels(P<0.01);the content of goethite and hematite decreased significantly(P<0.05).The changes in the relative contents of hematite and goethite corresponded to the changes in the red and yellow tones of the soil and reflected the differences in the moisture content of the soil environment.(2)Spectral characteristic bands of total iron,goethite and hematite in soilThe spectral curves of the soil samples are steeply cambered,and the spectra of total iron,goethite and hematite are significantly different in the 400 nm-600 nm band.The differential transformation and continuum removal treatment can effectively enhance the correlation coefficients.The continuum removal is most suitable for enhancing correlation coefficient between soil total iron content and reflectance.The second-order differential and logarithmic first-order differential transformation have the best effect on enhancing the correlation coefficient between goethite and hematite content and reflectance.The extreme values of the total iron content correlation coefficient were mainly located at 427 nm,518 nm,523 nm,559 nm,565 nm and 570nm,the spectral characteristic bands for the goethite content were 508 nm,617 nm,977 nm and 1011 nm,and the spectral characteristic bands for the hematite content were 442 nm,447 nm and 580 nm.(3)The hyperspectral inversion models of total iron content,goethite and hematite content in soilPartial least squares regression analysis is a more suitable hyperspectral modeling method for retrieving soil iron oxide content.The best hyperspectral inversion model for soil total iron content is a partial least squares regression model with reflectance continuum removal processing and constructed based on the full waveband,which was modeled to reach an R~2 of 0.9359,an RMSE of 3.42 g/kg,a predicted R~2 of 0.9818,an RMSE of 3.16 g/kg,and an RPD of 4.2352.The hyperspectral model with the best inversion of soil goethite content is a partial least squares regression model with a second-order differential treatment of reflectance and constructed based on the full wavelength band,which was modeled to reach an R~2 of0.9184,an RMSE of 0.67%,an RPD of 3.13,and a predicted R~2 of 0.9011 and an RMSE of 1.01%.The first-order differentiation of the logarithm of reflectance and the partial least squares regression model constructed on the basis of the full waveband are the best hyperspectral models for the inversion of soil hematite content,which is modeled to achieve an R~2 of 0.8247 and an RMSE of 0.33%,with a predicted R~2 of 0.6952 and an RMSE of 1.00%.(4)Technical process of hyperspectral inversion of total iron,goethite and hematite content in soilFrom the demand of non-destructive and rapid monitoring of soil iron oxide content,the technical process of hyperspectral inversion of soil total iron,goethite and hematite content was developed.Firstly,the soil reflectance was processed by continuum removal,second order differentiation and first order differential of logarithm,respectively,and then the sample set and validation set were divided according to the ratio of 0.75:0.25.Finally,a partial least squares regression inversion model is developed based on the full waveband.This paper clarified the vertical evolution characteristics of iron oxide in mountain soils of red and yellow loam areas,explored the spectral characteristics and hyperspectral inversion models of soil total iron and major iron oxide minerals(goethite and hematite),established a technical procedure for rapid and nondestructive inversion of soil iron oxide content.It provides a theoretical foundation and scientific basis for further exploring the inversion potential of hyperspectral remote sensing technology on soil iron oxide content and promoting numerical soil classification in red and yellow soil areas.