Investigation Of Heavy Metal Content In Lower Reaches Of Mulan Creek Using A Hyperspectral Remote Sensing Estimation Model

Soil is the basis of human life.However,with the further development of social industrialization,a large number of waste pollutants enter into the soil via various medium which make the soil polluted in different degrees.Among them,heavy metals are one of the most important pollutants.Hyperspectral remote sensing technology has become an important way to monitor soil heavy metals due to the characteristics of non-destruction,non-contact elements and wide range rapid monitoring.However,the existing researches about soil heavy metal estimation using hyperspectral remote sensing technology did not systematically consider and compare different soil modeling methods and soil spectral transformation methods.This paper took the lower reaches of Mulan Creek as research area,used a variety of spectral processing methods and modeling methods to estimate four soil heavy metal elements,which are Cr,Co,Cu and As.Then,four soil heavy metal elements modeling results were compared and analyzed in order to get the best spectral transformation methods and modeling methods in the study area.In addition,it provided certain reference value for rapidly detecting region heavy metal content.The main conclusion were as follows.(1)In the range of 400?2400 nm,the soil Cr and Co elements in the lower reaches of Mulan Creek,after a variety of spectral transformations,still had a large number of stable and continuous bands passing the 0.01 polarity test level.However,as to soil Cu and As elements,only after specific spectral transformation methods,there had few dispersive bands passing the 0.01 polarity test level.(2)The best spectral transformation methods of Cr,Co,Cu and As elements in lower reaches of Mulan Creek all had derivative transformation,which represented derivative transformation could enhance the spectrum response information to four heavy metals,thus improving four heavy metal elements estimation accuracy.In addition,reciprocal transformation,reciprocal logarithmic transformation,band depth and logarithmic reciprocal transformation could also to a certain extent improve the specific soil heavy metal element model accuracy.(3)The four soil heavy metal elements model estimation result in the lower reaches of Mulan Creek was Cr>Co>As>Cu,and for Cr and Co elements,the average modeling determination coefficients were 0.800 and 0.604,the average validation determination coefficients were 0.948 and 0.671 respectively,which was much higher than that of soil Cu and As elements.In addition,for soil Cr and Co elements,the selection of modeling methods had a relatively low impact on model accuracy,while the model accuracy of soil Cu and As elements were greatly influenced.In general,the modeling results of principal component regression,partial least squares regression and geographic weighted regression were similar,which could meet the study area soil Cr,Co and As elements hyperspectral retrieval requirements.However,for soil Cu element,the estimation result of multiple stepwise regression method was better than other methods.