Study On The Estimation Of Soil Organic Matter And Heavy Metal Contents Based On Spectral Absorption Characteristics

Black soil in Northeast China is the precious cultivated land resource,while unreasonable land utilization have led to the decline of soil quality.It is important for the sustainable development of soils to obtain the contents of the organic matter and heavy metals in soils timely and accurately.Heavy metals are scarce in soil,and their spectral absorption characteristics are not obvious.The existing estimation methods of the soil heavy metal mostly utilized statistical methods to select the characteristic bands,which cannot accurately explain the retrieval mechanism.In this paper,the absorption characteristics of soil organic matter and heavy metals are identified using the hyperspectral data,which provides the support for exploring the retrieval mechanism of soil heavy metals.The application of airborne hyperspectral remote sensing can realize the monitoring of soil components in large-scale continuous space.With an implementation example in Yitong County,the characteristic bands of organic matter and the retrieval mechanism of heavy metals(As,Pb,Cu and Ni)are explored based on laboratory spectra.The contents of organic matter and heavy metals in the soil of the study area were estimated by airborne hyperspectral image.The main work and conclusions of this paper are as follows:(1)Continuum removal was carried out on the laboratory spectra and nine spectral absorption characteristic parameters were extracted:D420,D480,D610,D700,D840,D1780,D2200,A2200 and lgA2200 etc..The correlation coefficients among the contents of organic matter,heavy metals and the spectra after continuum removal are obtained.It has been found that organic matter and heavy metals have a point of correlation mutation near 2200 nm,according to which the band ratio is calculated.Moreover,the result of univariate analysis shows that modeling of band ratio around 2200 nm achived the best performance.The multivariate analysis shows that using eleven features of absorption depth,absorption area and band ratio near 2200 nm offered the best performance on the estimation of the contents of organic matter and heavy metals.Least squares can be utilized to estimate organic matter and As using multiple parameters.To satisfy the estimation requirements,the non-linear models should be utilized with regard to Pb,Cu and Ni and the support vector machine method ahcive the best performance.(2)The characteristic bands of organic matter and the retrieval mechanism of As,Pb,Cu and Ni are analyzed based on laboratory spectral.It is concluded that 700 nm and 2200 nm spectrum are the characteristic bands for estimating the content of organic matter.As can be adsorbed by iron/manganese oxides,organic matter and clay minerals in soil.The adsorption of As by organic matter and clay minerals is stronger than the adsorption by iron/manganese oxides.Pb and Cu are mainly adsorbed by organic matter and clay minerals in soil while Ni is mainly adsorbed by iron/manganese oxides and organic matter in soil.(3)Continuum removal was utilized for the airborne spectra,and seven spectral absorption characteristic parameters were extracted:D600,D1300,D1630,D2000,D2200,A2200 and lgA2200 etc..Band ratios and highly correlated bands are selected as supplementary features.The combination of spectral absorption characteristic and bands is more accurate than using spectral absorption characteristic parameters or bands alone.Extremely Randomized Trees(ET)and Adaboost were introduced for regression and the fitting ability of these two methods were found to be stronger than the traditional classical methods.The optimal models of organic matter,Cu and Ni are based on ET,with~2are 0.6995,0.5787 and 0.6106,respectively.The optimal models of As and Pb are based on Adaboost,with~2are 0.6242 and 0.5214,respectively.The contents of organic matter and heavy metals in the whole study area are estimated using the optimal characteristics and models.It is found that the content of organic matter is high in farmland soil on the banks of the river.The contents of heavy metals in soils near mining areas,human activity areas and banks of rivers are higher.It means that the mining production and unreasonable utilization of land by humans have a significant impact on the content of organic matter and heavy metals in soil.