Risk Assessment Of Soil Pahs Pollution In Complex Urban Landscape Context And Spatial Mapping Study

Urban soils are essential natural resources in the process of urbanization,and they play a crucial role in urban construction as the primary medium for human activities.However,they are susceptible to the strong influence of human activities due to their complexity,strong spatial heterogeneity,and instability.PAHs are exogenous soil pollutants that pose a significant threat to the ecological environment and human health,increasing the risk of cancer.Soil PAHs content and spatial distribution are traditionally measured using laboratory methods,which are tedious and costly.Thus,it is important to develop a quick and accurate method for soil PAHs inversion.Visible-near-infrared spectroscopy provides a comprehensive reflection of the physical and chemical properties of soil,and the C-H bond stretching vibration in PAHs molecules generates spectral signals,which show specific absorption characteristics in the visible and near-infrared spectral regions.This provides a theoretical basis for soil spectral inversion of PAHs and a scientific basis for large-scale and wide-area inversion of soil PAHs contents.However,there is a lack of basic theoretical studies on the influence of wavelength,spectral resolution,and spectral response function on the accuracy of inversion results when using visible-near-infrared spectroscopy for soil PAHs inversion,especially on the use of remote sensing images for soil PAHs inversion.Therefore,this study takes Wuhan,a large representative city in central China,as the study area and addresses the PAHs pollution problem in the study area.Firstly,risk evaluation and analysis of PAHs pollution sources were performed by using toxicity equivalent and eigenratio methods;then a soil PAHs spectral inversion model based on traditional linear regression model(partial least squares)and machine learning model(support vector machine and random forest)was constructed with near-earth visible-near infrared spectra as the base data to evaluate the effects of pre-processing method,wavelength and spectral resolution on inversion accuracy The effects of pre-processing method,wavelength and spectral resolution on the inversion accuracy were evaluated.Finally,this study reconstruct the soil spectra with the help of spectral response functions,discuss the effects of simulated satellite remote sensing data and satellite remote sensing data on the inversion accuracy of soil PAHs,and evaluate the accuracy and feasibility of inversion mapping.The main conclusions are as follows:(1)The content of total PAHs in the soils of the study area ranged from 0.06 mg/kg to2.53 mg/kg,with a mean value of 0.54 mg/kg and a standard deviation of 0.53 mg/kg.According to the Dutch classification criteria for total PAHs contamination in soils,most of the soil sample sites were lightly contaminated with PAHs,and a few were moderately and heavily contamination.Ecological risk assessment of soil PAHs contamination was further conducted based on the toxic equivalent concentration of benzo[a]pyrene(Ba P).The results showed that 46.08% of the soil samples had Ba P TEQs exceeding the soil ecological risk standard of 0.03 mg/kg established in the Netherlands,indicating the potential ecological risk of PAHs contamination in soil.The results of source analysis of PAHs using the characteristic ratio method showed that the soil PAHs mainly came from a mixture of petroleum combustion sources and coal and biomass combustion sources,and further source analysis of PAHs using principal component analysis showed that the first source was mainly traffic and heavy industrial sources,with a pollution contribution of up to61.03%;the second source was mainly coal and biomass low temperature combustion sources,with a pollution contribution of up to 38.03%.The second source is mainly coal and biomass low-temperature combustion sources,and the pollution contribution rate can reach 38.97%.(2)Soil VIS-NIR spectral data were preprocessed using eight spectral preprocessing methods,and soil PAHs inversion models were developed using partial least squares,random forest and support vector machine algorithms,and the results showed that the highest cross-validation accuracy was achieved using smoothing plus first-order derivative transformation and modeled with random forest.Based on this,the original spectra were divided into visible band,near-infrared band,near-infrared long-wave and near-infrared short-wave according to the band range,as well as the spectral data with the spectral resolution resampled to 10 nm,20 nm,50 nm,100 nm,200 nm and 400 nm,and the first-order derivative processing was applied to the spectra to build the soil PAHs inversion model using random forest.The results showed that the soil NIR spectral data obtained the highest inversion accuracy with the highest accuracy of R-squared of 0.54 and RMSE of 0.38;using the spectral resolution of 10 nm had the highest inversion accuracy with the highest accuracy of R-squared of 0.51 and RMSE of 0.39,and the prediction accuracy showed a significant decreasing trend with the decrease of the resampled spectral resolution.(3)The spectral response functions of Landsat 8,Sentinel-2 and GF-5 were used to reconstruct the near-earth spectral data and establish a prediction model of soil PAHs based on partial least squares,random forest and support vector machine.The results showed that the reconstructed GF-5 simulated spectral data based on random forest had the highest prediction accuracy,followed by Sentinel-2 and the worst by Landsat 8.After further processing the remote sensing images of Landsat 8,Sentinel-2 and GF-5 in the study area,the corresponding bare soil spectral data were extracted and the established models were evaluated using the random forest algorithm.The accuracy of the model was evaluated by random forest algorithm,and the effects of the three remote sensing images on the inversion results of PAHs were compared.The results show that GF-5 remote sensing image has the best inversion accuracy with R-squared of 0.30 and RMSE of 0.52,followed by Sentinel-2with R-squared of 0.29 and RMSE of 0.54,and Landsat 8 with R-squared of 0.25 and RMSE of 0.51.The inversion results provide the possibility of predicting urban soil PAHs on a large scale.The three results consistently indicated that the high value aggregation of PAHs was in Jianghan District,Jiangan District,Qiaokou District and Qingshan District,which is highly consistent with their pollution sources and is crucial for the subsequent pollution prevention in the study area.