Hyperspectral vegetation indices to detect hydrocarbon pollution

Abstract:

This study assessed the effects of hydrocarbon pollution on three tropical forest areas and developed hyperspectral indices suitable to detect those effects. Field spectroscopy data collected in the Amazon forest were used to estimate the biophysical and biochemical parameters of vegetation by applying the inversion process of the PROSPECT model. Using band combinations of the complete spectrum, contour maps of the coefficient of determination (R 2) for foliar parameters were calculated, based on which, hyperspectral indices were derived. These indices were assessed against the medium resolution imaging spectrometer terrestrial chlorophyll index and applied to an Hyperion EO-1 hyperspectral satellite image. Based on the contour maps of R 2, two hyperspectral indices were defined, namely the ratio spectral index (RSI) and normalized difference spectral index (NDSI) indices. Higher correlations were observed for chlorophyll content from the SPAD-502 model (R 2=0.74 and R 2=0.73, respectively), followed by chlorophyll content from the PROSPECT model (R 2=0.59 and R 2=0.53, respectively). The new indices were applied to a Hyperion EO-1 satellite image in order to assess the chlorophyll content. In the pristine forest area sampled during fieldwork (site 3), the results revealed a strong correlation of 0.92 and 0.96 by RSI and NDSI, respectively, followed by the secondary nonpolluted forested areas (sites 2a and 2b) (R 2=0.76 and R 2=0.93, respectively). In the polluted secondary forest (site 1), the correlation of band combination was weak (R 2=0.10 and R 2=0.40, respectively). This suggests that the effects of hydrocarbon pollution in the tropical forest can be detected by Hyperion EO-1 satellite images by observing an increase of RED (reduce chlorophyll content) reflectance, associated to an increase in near infrared reflectance.

Año de publicación:

2020

Keywords:

• Hyperspectral satellite images
• Field spectroscopy
• Hydrocarbons pollution
• Amazon Forest
• ECUADOR

Fuente:

scopus