Hyperspectral signature analysis of three plant species to long-term hydrocarbon and heavy metal exposure

摘要

Recent studies aim to exploit vegetation hyperspectral signature as an indicator of pipeline leakages and natural oil seepages by detecting changes in reflectance induced by oil exposure. In order to assess the feasibility of the method at larger spatial scale, a study has been carried out in a greenhouse on two tropical (Cenchrus alopecuroides and Panicum virgatum) and a temperate (Rubus fruticosus) species. Plants were grown on contaminated soil during 130 days, with concentrations up to 4.5 and 36 g.kg~(-1) for heavy metals and C_(10)-C_(40) hydrocarbons respectively. Reflectance data (350-2500 nm) were acquired under artificial light from 1 to 60 days. All species showed an increase of reflectance in the visible (VIS, 400-750 nm) and short-wave infrared (SWIR, 1300-2500 nm) under experimental contaminants exposure. However, the responses were contrasted in the near-infrared (NIR, 750-1300 nm). 47 normalized vegetation indices were compared between treatments, and the most sensitive to contamination were retained. Same indices showed significant differences between treatments at leaf and plant scales. Indices related to plant pigments, plant water content and red-edge reflectance were particularly sensitive to soil contamination. In order to validate the selection of indices, hyperspectral measurements were performed outdoor at plant scale at the end of the experiment (130 days). Leaf samples were also collected for pigment analysis. Index selected at day 60 were still sensitive to soil contamination after 130 days. Significant changes in plant pigment composition were also observed. This study demonstrates the interest of hyperspectral data for oil exploration and environmental diagnosis.