Evaluating the feasibility of multitemporal hyperspectral remote sensing for monitoring bioremediation

• Published in: International journal of applied earth observation and geoinformation 2015-02, Vol.34, p.217-225
• Main Authors: Noomen, Marleen, Hakkarainen, Annika, van der Meijde, Mark, van der Werff, Harald
• Format: Article
• Language: English
• Subjects: Benzene; Bioremediation; Hyperspectral remote sensing; Multitemporal; Pollution; Red edge position
• ISSN: 1569-8432; 1872-826X
• DOI: 10.1016/j.jag.2014.08.016

•We evaluate multitemporal hyperspectral remote sensing for bioremediation.•We use a normalization procedure to compare red edge values of meadows over time.•The total accuracy of classifying all monitored locations was 65%.•The accuracy was 71% when only polluted locations were taken into account. In recent years, several studies focused on the detection of hydrocarbon pollution in the environment using hyperspectral remote sensing. Particularly the indirect detection of hydrocarbon pollution, using vegetation reflectance in the red edge region, has been studied extensively. Bioremediation is one of the methods that can be applied to clean up polluted sites. So far, there have been no studies on monitoring of bioremediation using (hyperspectral) remote sensing. This study evaluates the feasibility of hyperspectral remote sensing for monitoring the effect of bioremediation over time. Benzene leakage at connection points along a pipeline was monitored by comparing the red edge position (REP) in 2005 and 2008 using HyMap airborne hyperspectral images. REP values were normalized in order to enhance local variations caused by a change in benzene concentrations. 11 out of 17 locations were classified correctly as remediated, still polluted, or still clean, with a total accuracy of 65%. When only polluted locations that were remediated were taken into account, the (user's) accuracy was 71%.