Real-time in situ detection of petroleum hydrocarbon pollution in soils via a novel optical methodology.

Petroleum hydrocarbon (PHC) contamination in soils is considered one of the most serious problems currently, of which the detection and identification is a fairly significant but challenging work. Conventional methods to do such work usually need complex sample pretreatment, consume much time and fail to do the in-situ detection. This paper set out to create a novel systematic methodology to realize the goals accurately and efficiently. Based on laser-induced breakdown spectroscopy (LIBS) and self-improved machine learning methods, the innovative methodology only uses extremely simple devices to do the real-time in situ detection and identification work and even realize the quantitative analysis of pollution level accurately. In the study, clean soils mixed with petroleum were served as polluted samples, clean soils to be the blank group for comparison. Based on the elemental information from the spectra obtained by LIBS, machine learning methods were improved and helped optimized the algorithm to identify the PHC polluted soil samples for the first time. Furthermore, a novel model was designed to perform the quantitative analysis of the concentration of PHC pollution in soils, which can be applied to detect the degree of PHC contamination in soils accurately. Finally, the harmful volatile component of the PHC polluted soils was also successfully and identified despite its extremely minimal content in the air. The newly-designed methodology is novel and efficient, which has extensive application prospect in the real-time in situ detection of petroleum hydrocarbon pollution.