University of Twente Student Theses

This website will be unavailable due to maintenance December 1st between 8:00 and 12:00 CET.

On the feasibility of HyMap imagery to investigate hydrocarbon polluted areas over time for remediation purposes

Hakkarainen, Laura Annika (2010) On the feasibility of HyMap imagery to investigate hydrocarbon polluted areas over time for remediation purposes.

[img] PDF
Abstract:Hydrocarbons (HC) have become a severe environmental and economical problem due to increasing use of petroleum products causing vast number of HC contaminated sites all over the world. As organic pollutants, HCs have negative effect on humans and the environment. HC polluted sites should therefore be remediated as soon as possible. This unfortunately is not always possible, since clean up of such sites is expensive and may take several years to accomplish. In order to optimize the cleanup of such areas, non-destructive methods such as hyperspectral remote sensing (RS) have been recognized to be a potential tool for detecting and monitoring HC polluted areas. In the current study, the aim was to investigate the applicability of hyperspectral (RS) techniques to detect the change over HC polluted areas on two HyMap scenes near a leaking gas pipeline in the Northern Netherlands. Different spectral techniques, such as absorption feature analysis, continuum removal, first and second derivative analysis, existing HC and vegetation health indices, were investigated on the spectra collected from polluted and clean sites. A novel change detection method based on an image processing technique developed by van der Werff et al. (2008) was further applied, in order to detect the change over HC polluted sites throughout the whole study area. This technique allowed the comparison of two images automatically with normalized data, highlighting the variation of vegetation health status within the agricultural fields. The use of spatial factors such as pipeline maps further assisted on investigating vegetation stress related to only HCs. The results suggested that direct detection of HCs over non-vegetated areas was not possible for a number of reasons: for example the HCs from the leaking pipeline were either not situated on the surface or the remaining vegetation within the pixels hampered the detection of diagnostic HC absorption features. In contrast, using vegetation stress as indirect indicator of HC pollution, improved results were acquired: it was found that the Red Edge Position (REP) - index from several tested indices was the only index able to detect vegetation stress related to HC pollution, especially in densely vegetated areas with Normalized Difference Vegetation Index (NDVI) > 0.6. Lack of drilling data did not allow establishing spectra techniques for detection of HC pollution over partially vegetated sites (NDVI < 0.6). It was therefore recommended that vegetation indices taking into account the soil background reflectance (e.g. OSAVI, Optimized Soil Adjusted Vegetation Index), should be further investigated in detection of HC pollution over areas with less vegetation cover, in order not to leave any HC contaminated areas unnoticed. In the change detection analysis, several HC polluted locations were identified in which vegetation had become healthier between years 2005 and 2008. This indicated that HC pollution level had decreased on these sites, allowing improved growth of the vegetation. An overall accuracy of 58.8 % and Kappa coefficient of 0.13 was acquired for the change detection image. The accuracy assessment was however affected by the nature of the ground truth data (e.g. sampling depth) and therefore additional validation was recommended. This would not just verify the performance of the change detection method but also provide information if less cleanup actions are needed, further resulting in decrease of remediation costs in the study area.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page