Estimation of Pb concentration in the mining tailing areas base on field spectrometry and support vector machine

Heavy metal pollution of soil in the mining tailings is one of the most serious environmental problems in the world nowadays. Heavy metals may pose a long-term potential threat to ecosystems and human health. Therefore, it is significant to monitor heavy metal pollutions of soil in the mining tailings, and it become an urgent task. The goal of this research was to estimate Pb concentration in mining tailing areas base on field spectrometry and support vector machine. This research take Jinduicheng Mo mining tailings, Huaxian, Shaanxi Province as the study area. A total number of 288 soil samples collected at the mining tailing areas. The original dataset was divided into a training or calibration dataset (n=252) and an external validation dataset (n=36) for the prediction model. The soil samples were air dried and passed through a 2 mm sieve, then the heavy metal concentrations of Pb in soil were determined through chemistry analysis in the laboratory by graphite furnace atomic absorption spectrometry (GB/T17141-1997). The original reflectance spectral measurements of soil were collected using an ASD field spectrometer for the solar reflective wavelengths (350-2500 nm) in the field. The spectral measuring time is 10:30-12:00 with the outdoor natural light condition. The 8 ° angle probe was used, the distance from the contact probe to the surface of soil samples was set to 1.35 m in order to get the soil spectral in the range of 1m², and each soil sample was achieved 10 spectral measurements. The relationship between the spectral response of soil and Pb concentrations in the soil was determined. The original reflectance was transformed into a FirstDerivative Spectrum (FDS). Particle swarm optimization (PSO) were adopted to select parameters of support vector machine (SVM), then support vector machine was trained the training data set, in order to establish hyperspectral prediction model of Pb concentration in mine tailings soils. The hyperspectral- prediction model of Pb concentration in soils yielded a correlation coefficient of 0.7545 and Root Mean Square Error of the prediction (RMSE) value of 9.7612. The result indicates that the hyperspectral prediction model of Pb concentration in mine tailings soils based on support vector machine can predict Pb concentration rapidly, but the accuracy needs to be improved. The research will provide basic theory and method for predicting heavy metal of Pb in mine tailings soils, and will provide theoretical basis and technological support for controls of mining tailings and mining wasteland and its ecological restoration and reconstruction.