Broadband multispectral indices for remote sensing of vegetation affected by oil spills in the mangrove forest of the Niger Delta, Nigeria

Detection of vegetation affected by oil spills in oil polluted environments such as mangrove forest can be challenging using in-situ measurements and laboratory-based analysis techniques. Satellite remote sensing has been shown to be an effective tool to detect and monitor vegetation health and status in polluted areas. The application of broadband multispectral vegetation indices (BMVIs) derived from remotely sensed satellite data to detect and monitor impacts of oil spills on vegetation health has not been fully evaluated through previous research. The study was conducted in the mangrove forest South-West of Port Harcourt City in Niger Delta, Nigeria. This study first investigated the potential for using BMVIs to detect the impact / the effects of oil pollution on vegetation health. A total of 20 BMVIs were evaluated using data acquired at the visible, near infrared and shortwave infrared wavelengths. In Chapter 4 a statistical analysis of the indices from 37 oil polluted and non-polluted (control) sites show that 12 BMVIs demonstrated significant differences (p<0.05) between pre- and post-spill observations. For the control sites 11 of the 20 BMVI values did not indicate significant change and remained statistically invariant before and after the spill date (p ≥ 0.05). Oil spills are therefore suggested to cause a biophysical and biochemical alteration of the vegetation, leading to changes in reflectance signature detected by these indices. Five spectral indices (normalized difference vegetation index (NDVI), soil-adjusted vegetation index (SAVI), adjusted resistant vegetation index (ARVI2), green near infrared (G/NIR) and green shortwave infrared (G/SWIR)) were found to be consistently sensitive to the effects of oil pollution on vegetation and hence could be used for detection of oil pollution in vegetated areas. This study sought to, secondly, investigate factors that have been assumed to be influential on the detection of the impacts on vegetation from oil spills such as oil spill volume, time gap (number of days between oil spill events and image acquisition date) and spatial distance using the five BMVIs (NDVI, SAVI, ARVI2, G/NIR and G/SWIR). Regression analysis, utilised to determine the relative influence of these factors over 56 oil spill sites, revealed a significant relationship between the volume of the oil spill and increased deterioration of vegetation condition (p < 0.05) for four of the indices (NDVI, SAVI, ARVI2 and G/NIR). The length of time between image acquisition and oil spill was observed to exert an influence on the ability to detect the biophysical effects of oil spills on vegetation. The longer the time gap between the date of image acquisition and the oil spill event, the lower the detectability of oil spill impacts on vegetation. The influence of spatial variation on the detection of vegetation impacts was evaluated using a directional flow model applied over a local neighbourhood; the results from which did not show any significant difference between the neighbouring pixels (first pixel-P1, second pixel-P2 and third pixel-P3). The study also attempted to assess and validate the techniques used in chapter 4 in a different study site (study site 2- SS2) with a relative climatic and environmental conditions using new oil spill data in 2014. The findings revealed that statistical results from five indices (NDVI, SAVI, ARVI2 and G/NIR) derived from Landsat 8 in SS2 are found to show similar results to the ones obtained in SS1 using Landsat 5 & 7. In conclusion, it was found that the BMVIs have potential capacity for detection of vegetation affected by oil spills, not only are several factors found to exert a significant influence on the detection of oil spill impact on vegetation pollution using BMVIs, but also this method has the potential for replication in other over an oil-polluted environment.