Study On Hyperspectral Inversion Model Of Heavy Metals In Soils Of Shanghai Multi-scale City

The problem of heavy metal pollution in urban soil is one of the hotspots of social concern in recent years.Heavy metals in urban soil not only pollute the quality of the land,but also threaten the health of the residents through the atmosphere,water,food chain and other means.Investigating,evaluating,and rehabilitating heavy metal pollution in urban soils is an important part of ensuring urban land security and residents' safety.Under the background of rapid urbanization,urban soil is disturbed by strong human factors,soil structure and properties have changed to some extent,spatial heterogeneity of soil heavy metals has increased,and monitoring of heavy metals in soil has become more difficult.The traditional laboratory heavy metal detection method has high measurement accuracy but is time-consuming and laborious,and can only obtain instantaneous information within a limited range.The data has hysteresis and cannot meet the monitoring requirements of heavy metals in urban soils today.This paper selects the functional areas of Minhang District and Minhang District in Shanghai — industrial areas,traffic green areas and residential areas as research areas.The fractal method is used to analyze the spatial heterogeneity of soil heavy metal content,and it has high resolution and strong continuity.A wide range of hyperspectral remote sensing reflection data was used to model the linear and nonlinear inversion prediction of soil heavy metal content,and the influence of scale changes on soil hypermetallic hyperspectral inversion modeling was analyzed.The main findings are as follows:(1)Except for the heavy metal Cu in the residential area and Minhang District,the contents of heavy metals Cu,Cr,Mn,Pb and Zn in the other areas have higher than the background value of soil environment in Shanghai,and there is a certain degree of soil heavy metal enrichment.(2)The heavy metal Cr in the industrial area,the heavy metal Zn in the traffic greenbelt,the heavy metal Cu in the residential area,and the heavy metal Cu in the Minhang District have higher fractal dimension.The proportion of random components of these heavy metal spatial structures is higher.The spatial variation of random heavy components of soil heavy metals at different scales is affected by natural sources in the region,human activities such as traffic activities and domestic waste discharge.The stronger the human disturbance factor,the higher the fractal dimension and spatial structural features show higher randomness.(3)The overall trend of the spectral reflectance curves of soils at different scales is slowly inclined.Influenced by iron oxides,soil organic matter,and clay minerals,there is an absorption band in the spectral reflectance curve of the soil.The optimal feature bands at different scales located at different locations,and the range of feature bands is also different.(4)Nonlinear model for soil heavy metals Cu,Cr,Mn,Pb,Zn in Minhang District,heavy metals Cu,Cr in industrial areas,heavy metals Mn,Pb,Zn in traffic green soil,and inversion of soil heavy metals Cu,Cr and Zn in residential areas The prediction effect is better than the linear model;the linear model has better effect on the inversion of heavy metal Mn,Pb,Zn in the industrial area,heavy metal Cr in the traffic green space,and heavy metal Mn and Pb in the residential area.Linear and nonlinear models can meet the needs of soil heavy metal inversion prediction at different scales.(5)Low-order differential transformation of spectral reflectance can enhance the correlation between soil heavy metal content and reflection spectrum,and provide data reference for inversion modeling.As the scale shrinks,the correlation between soil heavy metal content and soil spectrum increases.The nonlinear model has a good predictive ability for soil heavy metals with high randomness of spatial structure.When the random factor of soil heavy metal spatial structure is relatively low,the linear model shows better inversion ability and soil heavy metal content.The stronger the correlation with the spectral reflectance,the higher the accuracy of the linear model inversion.