Mangrove spectra changes induced by oil spills monitored by image differencing of normalised indices: Tools to assist delimitation of impacted areas

Abstract

The delimitation of environmental impacts in estuaries caused by oil spills is a dynamic process. Delimitation of the impacted area based on the immediate adverse oil effects neglects responses that can manifest long after the spill has occurred. This study monitored two Brazilian estuaries damaged by oil spills, searching for their effect on mangrove vegetation. The spills spread large amounts of oil into the estuarine waters of Guanabara and Paranaguá Bays, affecting the mangrove and other environmental compartments. The mangrove regions in these bays were monitored under NDVI (Normalised Difference Vegetation Index) and NDII (Normalised Difference Infrared Index), using different epoch Landsat 5TM and 7ETM+ images contemporaneous to the spills, radiometrically corrected and coregistered. Differencing images (ΔNDVI and ΔNDII) were used to assess response to the spill. Differencing images identified areas where vegetation signals had decreased or increased. In order to track seasonal mangrove spectra variations, Reference areas were chosen in each studied bay. Mean pixel value (µ) and standard deviation (σ) were calculated from differencing images of said Reference areas and defined threshold values were applied to the Monitored areas. Differencing images of Monitored areas presenting pixel values below the lower threshold (µ − 3σ) suggested a significant decrease in the normalised index and oil-impacted mangrove vegetation, whereas pixel values above the upper threshold (µ + 3σ) suggested a significant increase in the normalised index and recovery of impacted mangrove. Results showed that 4.95 ha of Monitored areas 1 and 2 in Guanabara Bay presented a significant decrease in NDVI 40 days after the spill; 10.9% of which also presented a significant decrease in NDII. An area of 9 ha of Monitored areas 1 and 2 in Paranaguá Bay presented a significant decrease in NDVI 20 days after the spill; no area presented such signals before the spill and just 0.9 ha still presented pixel values below the lower threshold 7 months after the spill. An area of 8.64 ha of Monitored area 3 in Paranaguá Bay presented a significant decrease in NDVI 20 days after the spill, 3.24 ha of which showed significant increase in NDVI after 7 months, compatible with vegetation recovery. Artifact signals induced by tides were discussed. The results point to areas not considered in the official environmental assessments currently available, thus enabling an enlargement of the original impacted area delimitations. The methodology used is suitable for initial and rapid remote assessment, with the aim of defining hotspots for fieldwork examinations, such as those required during environmental forensic examinations.