Hyperspectral Remote Sensing for Detecting Vegetation Affected by Hydrocarbons in the Amazon Forest

This thesis seeks to understand the effects of hydrocarbons on the vegetation of tropical forests. It explores hyperspectral methods to detect changes in biophysical and biochemical parameters of vegetation affected by hydrocarbons in the Amazon rainforest of Ecuador. The literature review revealed that experiments in the laboratory, showed that in specific species hydrocarbons caused a reduced level of chlorophyll content, which is an indicator of stress. However, it was unclear whether the same effect would be observed in tropical forests. Fieldwork was conducted in several sites of the Amazon forest of Ecuador to establish whether this was the case. Foliage samples were collected in sites located within oil spills and also from pristine forest in the Yasuni National Park. More than 1,100 leaves from three different levels of the vertical canopy profile (upper, medium and understory) were analysed for biophysical, biochemical and spectral properties. A second-order polynomial chlorophyll content model was estimated based on several published calibrations models which use a portable chlorophyll meter. Modelled chlorophyll content showed high correlations with methods using reflectance indices (0.76) and the inversion process of the PROSPECT radiative transfer model (0.71). The analysis of biophysical and biochemical parameters at the three canopy levels of vegetation growing near hydrocarbons leakages showed decreasing levels of foliar chlorophyll content which suggest a reduced photosynthetic activity, higher levels of water content, which may explain the thicker leaves in the upper canopy, and thinner leaves in the understory. Based on these results, hyperspectral Hyperion and CHRIS-Proba satellite images were used to explore the potential of several vegetation indices to detect the symptoms of vegetation affected by hydrocarbons. The results indicated that a combination of an index sensitive to chlorophyll content at canopy level (Sum Green) with the NDVI index (Normalized Difference Vegetation Index) are suitable to detect vegetation affected by hydrocarbons. Those indices accurately identified vegetation growing near sites polluted by the petroleum industry and also when applied to an area affected by hydrocarbons from natural macro-seepages, and areas where hydrocarbons may be near the surface. Two new vegetation indices are proposed to identify vegetation affected by hydrocarbon pollution. Those indices showed sensitivity to differentiate secondary forest polluted and non-polluted. Chlorophyll content maps were computed based on an approach which uses the MTCI (MERIS Terrestrial Chlorophyll Index) at leaf level and scaled up to canopy level. The results of this research contribute to knowledge of regarding forest degradation. The approach could be used to detect hydrocarbon seepages as indicators of petroleum reservoirs, as well as significant pollution from oil spills in forest ecosystems. Moreover, the parameters for hydrocarbon stressed vegetation could be employed in a carbon cycle model to explore the impacts of hydrocarbon pollution on the carbon dioxide and water fluxes from tropical forests which are crucial for the carbon and water cycles.